Our study examined the associations between healthcare provider density and outcomes of care using data from a large nationally representative cohort of patients receiving cART in Uganda. Patient outcomes, as measured by loss to follow-up and mortality, were not significantly associated with the number of doctors, nurses, clinical officers, field officers, or healthcare providers overall.
Although healthcare provider density has been shown to influence patient outcomes and health indicators in cross-country examinations 
, this was not the case among patients attending TASO clinics in Uganda. It is likely, however, that the relationship between healthcare provider density and patient outcomes is linked closely with country- or community-level characteristics and cohort attributes. We cannot know the extent of generalizability of our study as cART programs in other countries infrequently provide outcomes on health worker status and clinical outcomes. A recent study by Lambdin et al, in Central Mozambique, found that patients attending clinics with larger patient volumes and a larger number of patients per pharmacy staff had a higher risk of patient attrition (defined as lost to follow-up or death) 
. This association, however, was not significant when the authors considered the number of patients per clinicians 
At TASO, it is probable that the density of volunteers available to patients is associated with outcomes of care. TASO uses a community-based model of support for patients receiving cART, where each patient has a supporter, often a family member or friend, who provides interpersonal support and reminds the patient to adhere to their cART regime. Given that the presence of such social supports is a widely accepted facilitator of positive patient outcomes 
, the presence of such a supporter may be a better predictor of patient outcome than the presence of formal healthcare providers in this setting. However, the TASO database does not include data on the density of community-based supporters, so its relationship to patient outcomes could not be examined.
The link between healthcare provider density and patient outcomes in the TASO setting are likely influenced by additional factors not captured in the database. For example, broad socioeconomic and political factors vary between sites. For example, during the study period, Northern Uganda was facing a civil war, particularly affecting the Gulu region 
. While evidence shows that HIV treatment can be effectively provided in such settings 
, variations in political and socioeconomic factors may influence healthcare provider retention and patient loss to follow-up and mortality. The interplay between these such factors with HIV service delivery is important and well accepted 
, but cannot be analysed given the data available.
Despite the lack of significant relationship between patient outcomes and healthcare provider density, the study also revealed that across service sites there exists variation among these variables. Across all ten sites, healthcare providers are in short supply and half of the sites report that unmet demand for healthcare workers is a key challenge to provision of care.
As with any study of this nature there are several limitations to consider. First, the TASO database does not include information on HIV viral load or HIV resistance, which would have been important indicators of HIV disease progression to consider. Also, it was not possible to include CD4 cell count as a patient outcome in this study because the TASO database does not include complete longitudinal CD4 cell count data. This lack of complete CD4 cell counts is a reflection of the diverse settings in which TASO works in Uganda. This problem is common in other resource-constrained settings as well 
. Our study did not demonstrate a relationship between the density of healthcare workers and either loss to follow up or mortality. It is possible that our study was hampered by the number of centres, rather than the number of patients, that may have restricted variance. It is possible that with more centres we would have observed a different finding.
We created indicators of healthcare providers and outcomes. Although density of health workers is a well-established indicator, there are both pros and cons to applying this as a rate. We considered all health cadres individually and also a pooled group. We have no doubt that some providers may have greater skills than others. Similarly, providing AIDS care may have been more complicated during the early days of cart provision (eg. 2004) when those initiated onto treatment were typically very sick. This may have changed over time. Also, our mortality rate addresses known deaths. Although TASO has a low rate of loss to follow up, we would expect that some patients that were lost had, in fact died, thus our mortality rates may underplay mortality rates. Because our study examined both attrition and mortality separately, and because TASO has active retention of patients, we believe our findings are broadly applicable. In studies examining mortality outcomes as a primary outcome, we typically assume that 50% of those lost to follow up have died.
The quality of the data and outcomes of this study are strengthened by the large sample size and the multiple years of data available for analysis. The cohort also represents cART patients nationwide, thus capturing a wide range of differing patient experiences and allowing for regional comparisons. TASO offers services across all regions of Uganda, though specific TASO centres or through Ministry of Health/TASO supported centres in more rural settings. While other providers are available in Uganda, the cART regimens offered are similar and the laboratory services offered are similar. Further, the use of community-based patient supporters to facilitate adherence, as the literature suggests 
, contributed to high adherence rates in TASO patients 
, a key determinant of patient outcomes. The TASO cART delivery model also incorporates active retention strategies to locate patients who do not attend their scheduled appointments, thus reducing the degree of loss to follow-up 
. In this cohort, 93% of patients were retained over the four years. In contrast, a systematic review of 32 ART programs in Africa showed that, on average, patient retention rates are 61.6% over a two year period 
Effectively providing HIV/AIDS treatment in resource-limited settings requires careful consideration of the health human resources, including the availability of healthcare providers. Policy or spending decisions in this area can have a direct impact on patient clinical outcomes. The results of the present study indicated that there were no significant associations between patient outcomes and healthcare provider density in Uganda. This suggests that that other factors, such as the presence of volunteer patient supports or broader political or socioeconomic influences, may be more closely associated with outcomes of care among patients on cART in this setting.