This, to our knowledge, is the first study regarding the association of structural and organizational factors with quality of HIV care in resource-limited settings. Specifically, we investigate the drivers for variation in adherence to laboratory testing protocol by health care workers across facilities. Adherence to these tests is critical to ensure good patient outcomes because WHO staging and CD4 count at initial visit are critical inputs to initiation of ART. In the presence of resource constraints, adherence to protocol and making right decisions also has implications at the population level. For instance, initiating an ineligible patient on ART without conducting all the tests also implies not initiating another eligible patient on ART. Similarly, not conducting certain tests for monitoring can have undesired consequences, especially if the frequency of these tests is low.
Our results indicate that health workers adhere to similar aspects of treatment protocol more strictly during initial visit than follow-up visits. This could be because the initial visit is more standardized whereas follow-up visits are more customized depending on the patient’s health status. Also, initial assessment can be construed as more critical since it determines treatment eligibility. Moreover, there are significant differences in the associations of organizational factors with adherence to protocol in initial versus follow-up visits.
The importance of physical space in ensuring quality of care in resource-limited settings has been mentioned before in the literature [20] and overcrowding has been cited as one of the primary drivers of inadequate care in emergency rooms [24, 25]. However, ours is the first study that provides empirical support for this notion. The procedure of taking blood requires time, space and privacy, which might explain why increased space was associated with higher odds of ordering and conducting repeat CD4 test. In addition, limited physical space creates a situation of crowding, convoluted patient flow, which can aggravate and confuse staff and patients leading to compromised quality of care.
Some seemingly counterintuitive findings could be related to the design of clinical protocols with some staff cadres permitted to perform specific activities but not others. For instance, nurses are required to double-check the ordering of blood tests during FUP visits to minimize the impact of clinician oversight. This potentially explains the findings that burnout increases the odds of conducting CD4 test during FUP visits. It is plausible that burnout amongst clinicians, who tend to be the most overloaded and therefore most stressed, may be (over) compensated for by less burnt out nursing staff [11].
Similarly, the result that greater staff experience decreased and higher staff turnover increased the odds of conducting tests appears counterintuitive from a developed country perspective. However, we believe that this is plausible in a high workload setting with protocolized care. For example, less experienced staff members tend to adhere better to protocols and newly introduced staff may want to demonstrate their performance to supervisors by being more compliant with protocols. This effect may wane over time as workers become complacent with attention to protocol details. More experienced staff and those who have worked at a clinic for more time may therefore put less emphasis on following protocol, relying more on personal experience or clinical judgment. This study suggests that turnover might be beneficial if it facilitates the replacement of demotivated and burnt out staff at the facility.
The results also demonstrate that some explanatory variables influence different outcomes differently. Due to the highly protocolized nature of ART and the step-by-step nature of healthcare delivery in this setting, responsibility for certain tasks often lies with different personnel in different parts of the clinic. One of the results of this style of care is that responsibility for certain tasks is atomized, making these ‘outcomes’ susceptible to different variables. For example, some outcomes are highly dependent on the staff member responsible for registering /enrolling clients, and that staff member is often different (including potentially having different training) to those responsible for WHO staging, or for those responsible for drawing blood to follow-up on test orders.
Another plausible reason for these different effects is that faced with time constraints, the staff members might prioritize their cognitive efforts on some tasks over others based on what they believe is more important. For instance, CD4 count at the time of enrolment is considered very important for all future treatment decisions. As a result, despite heavy workload and high levels of burnout, the staff members tend to not overlook this test.
Based on the authors’ programmatic experience, noncompliance to protocols is likely to be more common amongst Clinical Officers (COs) than nurses, due to the acute bottlenecks experienced at the point of patient screening, which promoted a culture of ‘clearing’ patients as quickly as possible. Nonetheless, nurses and COs working in ART maybe more compliant to protocols compared to other departments due to more a rigorous and continuous system of quality assurance checks.
The strengths of this study include the availability of electronic clinical and laboratory data on a large numbers of patient visits. We also had access to architectural data on physical space and information on levels of health care worker burnout, which was conducted in this setting.
However, there are limitations arising from the fact that the data were not originally collected for this study. The limited scope of the study prevented us from collecting data on additional measures pertinent to our objective. One such measure is leadership of nurse in-charge. Programmatic experience of authors in Zambia strongly suggests that facility-level leadership often plays an important role in adherence to protocolized care in weak health systems. However, creating a leadership index would require conducting a survey among nurses, which was beyond our scope. Future studies should develop and/or refine existing methodologies and collect prospective data to investigate this link.
Our measure of staffing was derived from administrative (payroll) database. This almost certainly results in some overestimation of the actual level of staffing in the clinic, due to unannounced absences from the facility and the practice of getting paid for the shifts but not being physically present in the clinic. Anecdotal evidence indicates that both practices were widely prevalent among clinic staff during the study period. Our measure of absenteeism was self-reported, based on the 2007 survey of healthcare workers and it did not account for unplanned absences. Our measure of experience (number of years in a particular grade) also partly captures the effect of age of the health care worker, particularly since the ART program started only 3–4 years before the survey. Thus, our results could be interpreted to imply that younger workers are more likely to adhere to the protocol because either they are open to clinic management ideas or because their education is structurally different than their old counterparts or because they are less tied to conventional ways of doing things.
There was limited variation on some of our predictor variables such as absenteeism, burnout and turnover, which may have limited the statistical power to detect significant effects. Similarly, almost all facilities performed very well on WHO staging with very little variation across facilities. These factors might explain the lack of statistically significant results for many associations. Also, the number of patients in WHO stage IV disease was very low across all facilities, potentially explaining the lack of significant associations for laboratory investigations done.
Without access to data on outcomes at the level of individual health care workers, we are unable to comment on what factors differentiate some workers from others in being more or less adherent to protocols. Moreover, we note that the results cannot be used to assess the absolute quality of care provided at the study facilities. This is because of the lack of external benchmark on what constitutes good quality care in these resource-limited settings and because the study was not designed to answer this question. Because of imperfect patient adherence to visits, we had to allow sufficient time buffer (+/− 4 weeks) in constructing our quality measures.
Our choice of urban facilities in Lusaka, Zambia’s capital, limits the generalizability of our results in other locations, especially rural regions of the country. However, these results suggest similar assessments in other resource-limited settings attempting rapid scale-up of HIV care and treatment are necessary (particularly in sub-Saharan Africa) to ensure that the relationship between structure, process and outcomes in settings utilizing protocolized healthcare delivery are understood adequately by policy makers, donors and implementers.