The first few months following initiation of cART is a critical time for severely immune-suppressed HIV-infected patients. These data suggest that more frequent monitoring of patients in the early months by a dedicated nurse can significantly improve survival and retention in care among these high-risk patients and improve their retention in care. Although further evaluation is needed, this intervention may be relatively easy to implement in other resource-constrained environments.
To our knowledge, the concept of frequent nurse-based rapid assessments is among the few interventions other than cotrimoxazole prophylaxis that has been associated with a profound reduction in early mortality among high-risk HIV-infected patients initiating cART in low-income settings [
13,
14]. We believe the effects of HREC are a combination of the rapid
and frequent assessments. Rapid because this makes accessing healthcare more accessible to patients (by not having to wait as long and by not having to spend as much time in the clinic); frequent because it makes it more likely that early warning symptoms (e.g., fever, rash) can be identified within days, as opposed to weeks, of their onset. If, for example, the monthly standard of care visits were simply made more rapid, such symptoms as fever or rash would go unattended for potentially weeks, thereby increasing the risk of full-blown immune-reconstitution disease, more severe toxicities, etc.
We also postulate that more frequent monitoring is effective at improving early patient outcomes through both direct and indirect mechanisms. For example, earlier identification of the signs and symptoms of drug toxicity, opportunistic infections and immune reconstitution syndrome likely leads to earlier interventions to address these issues; thus patients should experience a direct survival advantage in the short-term.
Indirectly, HREC may have improved adherence to cART and thus improved short-term outcomes. Adherence may be improved in the HREC population because of the weekly contact, reminders and supports; as a result of improved adherence, patients will be more likely to experience complete virologic suppression, have improved immune-response, and be less likely to develop resistance, therefore indirectly contributing to survival benefits over the short and long term. HREC may have improved retention by enabling patients to be seen quickly without having to wait in long queues and without having to pass through multiple stations (i.e., spending much of the day at the clinic), thereby making their healthcare more accessible to them.
There are some additional costs associated with HREC. For example, there are added direct costs to patients arising from increased transportation required to and from the clinic. They may also have to miss more work because of more frequent clinic visits, translating into increased opportunity costs. The nurses hired to work on HREC were hired explicitly for that purpose, and this certainly adds to the overall programme expense on personnel. Whether these additional costs and expenses are justified when weighed against the costs associated with increased morbidity, mortality and loss to follow up is the subject for a detailed cost-effectiveness analysis that is beyond the scope of the present evaluation.
There are two key strengths to these findings. First, the standard of care in AMPATH is already to provide cotrimoxazole or dapsone until a patient's CD4 count is above 200 cells/mm3. These data have therefore been able to assess the effect of more frequent monitoring in a setting where the vast majority of patients were already receiving cotrimoxazole or dapsone. Second, the effect of HREC was strong across three different versions of the primary outcome, suggesting a robust effect. By evaluating the impact of HREC on both mortality and LTFU, we took account of the two crucial factors determining the success of an HIV treatment programme: keeping patients alive and in care. Moreover, we used statistical methods that appropriately and robustly adjusted for measured confounders.
There are also limitations to this analysis. First, the choice of clinics in which HREC was rolled out to first may have created a variety of potential biases in our analysis related to a possible higher quality of care offered at those clinics. For example, this may have created some selection bias (improved patient outcomes at those clinics irrespective of HREC, e.g., because of lower patient volumes, or higher functioning staff), and ascertainment bias (better ascertainment of death and other clinical outcomes at the higher functioning clinics). Similarly, although all patients were eligible to be enrolled into HREC, only a small proportion were enrolled. If the providers who were more likely to refer patients to HREC were also the ones more likely to be current with clinical protocols (i.e., to prescribe cotrimoxazole) and/or more likely to be better clinicians, those patients referred by them may have been more likely to have better outcomes anyway.
However, for those patients referred, it was the HREC nurse who had the majority of clinical contact with them, thereby reducing any potential provider bias on the part of the referring clinician. Similarly, a slightly smaller proportion of patients enrolled into HREC were WHO Stage III/IV at treatment initiation. Although these issues may have biased the findings favourably towards HREC, the use of propensity score methods helps to overcome this possible bias because weighting in inverse proportion to the treatment propensity score creates a pseudo-sample wherein allocation to HREC is independent of the confounders that have been included in the propensity score. Hence the weighted dataset can be analyzed as if the group allocation were random. A limitation of this method is that there may be non-random allocation to HREC based on unmeasured factors to the extent that receipt of HREC depends on unmeasured factors. We acknowledge that residual bias may therefore remain due to potential unmeasured confounders, including adherence to cART (not accounted for in this analysis due to unreliability of the data).
1,2,3,4 Abraham Siika,1,2,3 Joseph Hogan,3,5 Rose Kosgei,3 Edwin Sang,3 John Sidle,1,2,3 Kara Wools-Kaloustian,1,2,3 Alfred Keter,3 Joseph Mamlin,1,2,3 and Sylvester Kimaiyo2,3
This article has been 
