A Quality Improvement Intervention to Inform Scale-Up of Integrated HIV-TB Services: Lessons Learned From KwaZulu-Natal, South Africa

Lessons Learned

In the SUTHI trial, we identified important factors that may explain the suboptimal improvement for some indicators. The effect of baseline performance was to the advantage of IPT improvement; however, HIV testing services and ART initiation in HIV-TB coinfected patients were high at baseline, and there was little room for improvement thereafter. Future QI interventions should consider baseline performance when setting expectations for improvement, however, we do not recommend that baseline performance be considered the sole criteria for selecting indicators for improvement. This study showed that there are indicators that are close to reaching targets but appear to be plateauing, for example, VL testing (Figure 3B) and TB screening (Figure 3C). QI improved both indicators and still has a role to play in addressing the barriers that prevent these indicators from reaching the desired target of 90%.

Capacitating clinic staff with data analytic skills is an important factor in ensuring the success of QI interventions because it improves technical skills, confidence, and self-efficacy of clinic teams.³² In addition, monitoring improvement using routine data is fundamental to the QI intervention. Learning sessions covered the basics of how routine data can be analyzed (e.g., calculation of percentages, means, and medians), plotted onto run charts, and interpreted using run-chart rules. QI mentors reinforced this knowledge at QI mentorship visits. Poor data quality threatens clinic teams’ efforts to monitor improvement and is a barrier to successful QI implementation.³³ Despite our attempts to address the completeness and accuracy of routine data, poor data quality undermined our QI intervention. For example, TB screening data were adversely impacted by inflated clinic headcount numbers (the denominator), incorrect completion of TB screening registers, and misplaced TB registers. The success of VL monitoring improvement depended on accurate and complete data entered into the patient electronic database; however, nearly 6 months of addressing data entry backlogs reduced the time available to improve the indicator coupled with challenges of tracing of patients to return to the clinic. Tracking patients is a resource-intensive effort due to poor telephonic services, lack of vehicles, and incomplete patient contact information (namely, telephone/mobile data, lack of street addresses).

A Ugandan-based QI project that aimed to improve TB case notification also relied on routine data to monitor improvement and went beyond checking clinic registers for completeness and accuracy.³⁴ A data tool was used to triangulate patient data from multiple sources, that is, laboratory data, patients’ chart notes, and TB laboratory data.³⁴ Unfortunately, no data metrics were available to quantify the extent to which data was improved. A South African QI project to prevent mother-to-child transmission of HIV in labor wards used a specially designed checklist that included prompts for nurses to complete and document important tasks.¹⁷ Following this intervention, there was a marked improvement in data quality with erroneous data, namely, percentages greater than 100% being eliminated. These studies demonstrate that innovative measures are needed to improve the quality of routine data and adding additional human resources to improve the completeness and accuracy of data may not be sufficient.

We intended for PHC clinic supervisors to lead the QI intervention, but their involvement was limited by their workloads and conflicting meetings. Clinic staff selected to attend the learning sessions did not always pass on their learnings from the workshops to their colleagues and the study-appointed nurse mentors reported resistance from non-workshop attendees to the workflow changes. A mixed-methods study identified personal- and work-environment-related factors that influence a health care worker’s ability to transfer knowledge from QI trainings to peers.³⁵ Health care workers that are successful in transferring training knowledge have a positive attitude to implementing changes, interpersonal skills to address resistance from peers, and the ability to question the status quo.³⁵ A work environment in which teams are receptive to new ideas, supportive of change, and leadership support is present, facilitates the transfer of training knowledge. ³⁵ In the SUTHI trial, selection of clinic staff to attend learning sessions was at the discretion of the PHC clinic nurse supervisor and nurse in charge of the clinic. While individuals from all clinic departments were trained, staff categories, such as data capturers and lay counselors may not have been empowered enough to transfer their new knowledge to more senior colleagues. Future QI interventions must consider screening potential QI trainees for the appropriate qualities that will allow for the transfer of QI knowledge.

Challenges in QI Implementation

Implementation of QI at the clinic level was accompanied by several challenges. First, QI was vaguely understood in both districts and clinic teams often believed that they were implementing QI by virtue of the weekly nerve center meetings and discussing problems and challenges at staff meetings. The need for the SUTHI QI intervention was initially unclear to QI clinics. The learning sessions established the importance of using a QI approach that is guided by a framework (Model for Improvement and PDSA), uses tools (e.g., process charts), locally developed strategies (change ideas), and monitoring progress with data. Importantly, the consistent visits and mentorship by the QI nurse mentors were critical in demonstrating how the frameworks and tools translated to practice.

Secondly, QI implementation adds additional work for clinic staff, in that data needed to be collected and recorded to track progress more frequently. While change ideas were implemented, it was a challenge to keep staff motivated to track their performance. For example, in HTS, group pretest counseling was a key change idea; however, the source documents developed to track the number of group pretest counseling sessions in patients’ waiting areas were not completed.

Third, leadership at the clinic level was supportive of the QI intervention; however, due to many commitments in and outside of the clinic, there was little involvement of clinic leaders in the QI meetings. This delayed implementation of some change ideas, as junior-level clinic staff do not have major decision-making power to make changes, such as in clinic patient flow.

Limitations

The study had limitations. First, while the QI intervention was implemented in the context of a randomized controlled trial, we were unable to prevent exposure of QI clinics from other improvement interventions to enhance integrated HIV-TB services, particularly, improvement efforts of the DMTs and technical assistance from local NGOs. Motivated DMTs in both study districts frequently monitored the progress of HIV and TB process indicators, fed back to poorly performing clinics, and conducted quarterly in-person visits to all clinics. The true effect of the QI intervention has likely been masked by the improvement efforts of the DMTs and local NGOs. Although the study was unable to separate the effect of the DMTs’ efforts and QI intervention efforts on improvements observed, the baseline period (Figure 3A–3E) offers some insight into the performance before and after the QI intervention was implemented. The QI intervention ideally complemented the performance monitoring and feedback efforts of the DMTs which were seldom able to conduct in-depth root cause analyses of systems weaknesses and develop clinic-specific interventions. Second, the study was not adequately resourced to determine if improvements in the QI clinics were sustained beyond the study period or if the QI tools, strategies, and best practices were scaled up to more clinics in other areas. Staff turnover and changes in key personnel, who were trained in QI methods, may add to the challenge of sustaining and scaling up QI activities once the study ended. Third, as per the study design, all analyses were at the cluster level and clinics within each cluster were considered as 1 unit. However, the QI intervention was at the clinic level, and different clinics within a cluster adopted different change ideas (such as the different timing of IPT initiations in Table 4), and we could not compare clinics to determine which change ideas translated to larger improvements.