Evaluation of a training intervention to improve cancer care in Zimbabwe: Strategies to Improve Kaposi Sarcoma Outcomes (SIKO), a prospective community-based stepped-wedge cluster randomized trial

Abstract

Introduction: Most Zimbabweans access medical care through tiered health systems. In 2013, HIV care was decentralized to primary care clinics; while oncology care remained centralized. Most persons in Zimbabwe with Kaposi sarcoma (KS) are diagnosed late in their disease, and the prognosis is poor. Little is known about whether educational interventions could improve KS outcomes in these settings.

Methods: Interventions to improve KS detection and management were evaluated at eight Zimbabwe primary care sites (four rural/four urban) that provided HIV care. Interventions included a standardized KS clinical evaluation tool, palliative care integration, standardized treatment and improved consultative services. Interventions were implemented between February 2013 and January 2016 using a randomized stepped-wedge cluster design. Sites were monitored for KS diagnosis rates and KS outcomes, including early diagnosis (T0 vs. T1 tumour stage), participant retention and mortality. Analyses controlled for within-clinic correlations.

Results: A total of 1102 persons with suspected KS (96% HIV positive) were enrolled: 47% incident (new diagnosis), 20% prevalent (previous diagnosis) and 33% determined as not KS. Early (T0) diagnosis increased post-intervention, though not significant statistically (adjusted odds ratio [aOR] = 1.48 [95% confidence interval (95% CI): 0.66-3.79], p = 0.37). New KS diagnosis rates increased 103% (95% CI: 11-273%), p = 0.02) post-intervention; although paired with an increased odds of incorrectly diagnosing KS (aOR = 2.08 [95% CI: 0.33-3.24], p = 0.001). Post-intervention, non-significant decreases in 90-day return rates (adjusted hazard ratio [aHR] = 0.69 [95% CI: 0.38-1.45], p = 0.21) and survival (aHR = 1.36 [95% CI: 0.85-2.20], p = 0.20) were estimated.

Conclusions: KS training interventions at urban and rural Zimbabwe decentralized primary care clinics significantly increased overall and incorrect KS diagnosis rates, but not early KS diagnosis rates, 90-day return rates or survival.

Trial registration: ClinicalTrials.gov NCT01764360.

Keywords: HIV; KS; Kaposi sarcoma; palliative care; primary community care; training intervention tools.

Figure 1

Timeline for study implementation and completion for randomized stepped‐wedge cluster trial. Each site was monitored for Kaposi sarcoma diagnoses and outcomes throughout the entire study period (evaluation period, weeks 0–150). The light‐shaded area shows the time when each of the eight sites was monitored prior to implementation of the Intervention Package (pre‐intervention period). The dark‐shaded area shows the time when sites were monitored after the implementation of the Intervention Package (intervention period). The time of intervention implementation was randomly assigned for each site; because the Urban‐1, Urban‐2 and Urban‐4 sites are in close geographic proximity, and share staff and patients, these three sites were randomized as a cluster. The Urban‐3 site was the first site to begin the intervention at week 15. The last site to begin the intervention was the Rural‐1 site at week 64.

Figure 2

Effect of the training intervention on Kaposi sarcoma (KS) diagnosis rate. The proportion of all weekly HIV clinic visits that were patients with a suspected KS diagnosis is shown for the duration of the study period for each site. Vertical dashed lines indicate the time that the training intervention was introduced. Sites are grouped by urban (left panel) and rural locations (right panel). The tick mark rug indicates study enrolment times for confirmed KS cases (black tick marks) and participants initially identified as having KS but later determined to not have KS by expert opinion (grey tick marks).

Figure 3

Diagram for identification of Kaposi sarcoma (KS) cases. A total of 1102 suspected cases of KS were evaluated during the combined pre‐intervention and intervention periods: 358 were determined to not be KS by expert opinion. Of the 744 confirmed KS cases, in 224 cases, the diagnosis of KS was made prior to study week 0. Of the 520 confirmed KS cases diagnosed after week 0, 74 were diagnosed during the pre‐intervention period and 446 during the intervention period.

Figure 4

Kaposi sarcoma (KS) diagnoses relative to the time of implementation of the intervention for each clinic. Time of diagnosis is shown for the 520 new confirmed KS cases relative to the time of the intervention at each site (vertical shaded line). Filled circles are ACTG stage T0 KS; empty circles are ACTG stage T1. The proportion of T0 diagnosis in each period is shown.

Figure 5

Adjusted Cox proportional hazards model of survival. Adjusted Cox proportional hazards model of survival of new (incident) Kaposi sarcoma (KS) cases by pre‐ and post‐intervention status (denoted by line width) with stratification by enrolment at a rural (vs. urban) clinic and tuberculosis status (denoted by colour). Adjustment covariates age, sex and time since HIV diagnosis were set to median values (age = 37 years, sex = male, time from HIV to KS = 0.92 years) with separate graphs for T0 and T1 status. Only the 520 confirmed incident (newly diagnosed) KS cases were included in the analysis. Censoring occurred at the maximum clinic visit.