For too long, debates around the positive versus negative synergies and systemwide effects of HIV funding on other programme and disease control priorities, especially in sub-Saharan Africa with its multiple burdens of disease and fragile health systems, have ignored the need for evidence, or have relied on key informant interview studies . These are valid methods, which we have used, but are not the method of choice for quantifying and comparing client numbers and trends for HIV and non-HIV priority services. They are vulnerable to bias, or at least wishful thinking, in the hotly contested arena where evidence of positive or negative effects of earmarked funding for specific diseases could influence donors' funding decisions. This paper does not provide the answer to the question: is successful scale-up of HIV services, which this paper confirms is happening in Zambia, having a beneficial or detrimental effect on other priority services. However, it does provide an answer, based on a small sample of facilities; and - more importantly - it provides an evaluation approach that needs to be replicated in larger samples and preferably on a national scale.
In facilities that had begun to rapidly scale-up HIV services and continued to provide other priority mother and child health services, the findings provide some support for the view that the former was having positive effects on the latter, at least with respect to antenatal care and family planning. How reliable is such evidence and how much weight can be placed on it? Firstly, there is a plausibility to it, or at least plausible hypotheses can be drawn from the findings, which need to be explored or tested. For example, there are obvious reciprocal connections, for example between PMTCT and antenatal care: one would expect (or at least want) that a facility that is offering PMTCT, as part of its antenatal care, would attract more pregnant women to attend for antenatal care. Also, antenatal care of women who come to a health facility when pregnant, is the main route to PMTCT. The linkage between PMTCT and antenatal care in Zambia has been demonstrated by Potter et al (2008) , who demonstrated improved quality of antenatal care in Lusaka facilities delivering PMTCT, as evidenced by trends in documented syphilis screening in pregnant women.
The strong positive associations of family planning client numbers with PMTCT, ART and VCT client numbers are interesting. Integrated services for family planning and HIV have been promoted over recent years [21–23]. A systematic review, which mainly focused on project-type interventions that sought to link family planning with VCT services, concluded that the evidence demonstrated the feasibility and effectiveness of integration . The linkages are plausible and consistent with a move to provide integrated services to meet women's reproductive health needs, which dates back at least to the 1994 Cairo Conference on Population and Development.
We did not explore internal referral processes in these facilities, although others have reported on a programme that used antenatal care as an opportunity to initiate eligible women on ART . One would expect (or at least want) that women who register for ART, perhaps postnatally on completion of PMTCT, would be referred to register for family planning services. This was the strongest correlation (Spearman coefficient = 0.83) in our analysis. One study estimated that the impact of PMTCT services would double in 14 high prevalence countries, by integrating family planning and PMTCT services . A somewhat surprising finding in our analysis, which needs to be assessed in larger sample size quantitative analyses and explored in future explanatory mixed methods research (see Conclusion), was that the positive rank correlations of family planning client numbers with HIV service trends was partly due to downward trends in both sets of services in a minority of facilities and that numbers of family planning clients were increasing in facilities not delivering ART.
The significance of the study findings is that a population-based approach was used to select health facilities for the study. We did not select project facilities where a special primary health care approach was being used to ensure non-HIV priorities benefited in tandem with HIV service scale up. Nor was the study restricted to 'research facilities', although journal articles based on data from Lusaka indicate that a lot of ART and PMTCT research was going on in facilities that were included in our study [15, 25]. Three of 11 PMTCT - antenatal care facility pairings in our study were from facilities outside of Lusaka, as were 4 of 12 PMTCT-family planning facility pairings. While enhanced integration may have been a programme effect, one might also expect the opposite where a programme supported by PEPFAR funding would prioritise HIV services at the expense of family planning services. A contextual factor that would help explain integrated services and positive correlations is Zambia's policy of "integration and scaling-up of the Prevention of Mother-to-Child Transmission of HIV and AIDS strategy into maternal and child health services" as a strategy to reach Millennium Development Goals (MDG) 4 and 5 .
There are some caveats or limitations to this study, before considering its broader significance. Firstly, it is based on a sample of three districts, though because of the inclusion of the capital city and another urban setting where early HIV scale-up was taking place, it did capture a high proportion of national ART scale up between 2004 and 2007. Also, only facilities that were identified in a May 2008 mapping exercise as providing ART were included, which would have excluded facilities that began to deliver ART subsequently . Therefore representativeness cannot be assumed. However, it is the internal validity that is important, i.e. the findings do illustrate what was happening within the surveyed facilities.
The subsets of facilities where data were available to enable correlations to be measured resulted in small numbers (between 9 and 12 pairs for each analysis - see Table 3). The main reason for this was that the study aim was to demonstrate trends over time and the initiation of ART and PMTCT in 2006 in rural Mumbwa district meant that all of these facilities were excluded from the 2005-07 trend analysis. Consequently, Lusaka contributed most of the facilities to the analyses shown in Table 3 and Figure 3, ranging from eight of nine (for family planning and ART) to six of eleven facilities (for VCT and family planning). Hence, any hypotheses and assumptions around positive effects of HIV scale-up on other services have quite limited generalisability. Studies that utilise this approach need to be replicated with larger numbers and representative samples of facilities, capturing later years when HIV service roll-out has extended to more districts.
Secondly, routine health information collected by health workers has several limitations, though we would contend that analysis of such data is more important in the long run than conducting special surveys. Recording errors are possible at all stages from initial completion of reporting forms and client registers through to reporting at the national level, or as part of a research study [28, 29]. Such problems have been identified in Zambia  and in our study four facilities reported different returns for numbers of clients registered for antenatal care and six for family planning, when comparing two sources of data - facility record reviews and district office data. Such errors reflect the lack of attention there has been to health facility data and reporting, while acknowledging that steps to improve routine data collection have been included in Zambia's 2009-2015 national information strategy .
Immunisation data returns are normally considered to be inherently inferior to population-based data, in that the former are based on numbers and types of vaccine antigens delivered, rather than on numbers and ages of children immunised. Coverage rates are then based on estimates of the target population, rather than on actual children identified in household surveys. The denominators (catchment populations) can be an underestimate of the size of the target group because children from other catchment areas that lack immunisation services may travel to facilities that offer such services. This can result in numerators (numbers of children immunised) that exceed the denominators (catchment target population sizes), as was found in this study. However, despite difficulties in interpretation that are commonly found when analysing routine vaccination data returns, these findings provide evidence that district immunisation programmes have continued to be delivered at a sustained level to the populations in the catchment areas of the three districts where HIV/AIDS scale up has been taking place. Intra-facility analysis did not demonstrate any consistent correlation between HIV and immunisation service trends.
The primary aim of the studies conducted under the umbrella of the Global HIV/AIDS Initiative Network (GHIN) was to assess the wider systems' effects of global HIV funding. The GHIN researchers recognised from the start the importance of measuring and acknowledging the primary aim and effects of the GHIs, which was to scale-up and reach more people with the HIV interventions that they needed. The findings on HIV scale up (Tables 1 and 2, and Figures 1 and 2) correspond with and confirm those reported by other researchers in Zambia. Our PMTCT data showed a similar trend to that of Stringer et al (2006) , which reported a decline in HIV positivity from 25.7% to 21.8% among pregnant women who were tested. This was not surprising, given that there would have been overlap in the data sources in Lusaka, which provided 81.4% of the pregnant women that were HIV tested in our study.
One of the GHIN objectives was also to assess access and equity effects, for example to assess if rural dwellers were also benefiting from interventions that were first rolled out in capital cities and other urban centres. In that respect, while Mumbwa rural district is not as difficult to reach as many parts of the Northern Province of Zambia, the findings show that once significant scale-up had started in Mumbwa (PMTCT in 2006 and ART in 2007) client numbers and coverage rose rapidly.
The wider significance of the findings in this study is that they illustrate the potential to derive useful evidence - for district as well as national programme managers - from routinely collected health facility data. Two approaches to collecting evidence on the performance of health systems have dominated in the last 10 years, one top-down and the other bottom- (or population-) up: the Health Metrics Network has carried out valuable work on the development of indicators and the Institute of Health Metrics has demonstrated the power of collating data and comparing performance across countries and regions [29, 32]. Countries, such as Zambia, have followed this lead and have focused much of their efforts on aggregating data nationally and reporting to international fora such as the UNGASS .
The second approach has been bottom-up, where considerable efforts and funding has been allocated by donors to conducting household surveys, which - unlike health facility data - provide evidence of unmet need and health seeking behaviour, as well some evidence of services accessed [33, 34]. However, as has been recognised by some commentators [30, 35], insufficient attention has been paid to what is happening in the middle, that is at health facilities where disaggregated data need to be collected on performance, so as to identify good and poor performers and take action. In our initial 2007 survey, much effort was expended on collecting data on routine services (family planning, immunisations, antenatal care) directly from health facilities. Not uncommonly, data were missing because facilities had made data returns to the district office but had not retained copies at the health facility. This is symptomatic of what is fundamentally wrong about how health information systems (HIS) work in sub-Saharan Africa; or rather how they sometimes work for higher level planners (national, provincial and sometimes district), and do not work where they should work, which is at the health facility level.
In the 2008 survey we found that most of the routine non-HIV service data was available in disaggregated formats (disaggregated to individual health facilities) at district health offices. The non-HIV priority service data used in the paper were part of a functioning - if neglected - health information system. The senior researchers who were supervising the field work found little or no evidence that data were being analysed and acted upon at the facility, or even at the district level. One explanation for the non-use of data is the multiple burden of data collection at facility and district level in Zambia, due to parallel health information systems established to meet the information needs of global initiatives .
This paper demonstrates the feasibility of obtaining and analysing routinely collected data to illustrate the performance of non-HIV priority services in facilities where HIV services are scaling-up. Some findings should concern programme managers who have overall responsbility for the health services system. The reported availability in the previous 12 months of essential drugs for national priorities - malaria, bacterial infections and management of normal labor and obstetrical emergencies - was significantly poorer than for ARVs. While stock-outs were more frequent in the rural district, they were also (surprisingly) common in Lusaka. This suggests that, while AIDS funding can strengthen pharmaceutical management , non-HIV drug and commodity management had been less reliable in the previous year than it was for HIV. However, in the absence of trend data, one cannot infer any association between HIV and non-HIV pharmaceutical management.
The findings demonstrate upward trends in client numbers for non-HIV maternal and child health programmes, in three districts where HIV services were scaling-up. The analysis also demonstrates scale-up in reproductive health service client numbers (family planning and antenatal care), generally in the same facilities where HIV services were scaling up; and the rank correlation supports the interpretation that this was an aetiological link. However, the stated caveats mean that those who are seeking definitive evidence to conclude that investments in HIV benefit other service priorities should await more conclusive evidence. Interestingly, district childhood immunisations increased overall, but not in the facility catchment areas where HIV scale-up was happening.