Methods: Sociodemographic, clinical, and attendance data from patients with appointments (6, 12, 24, or 36 month appointments) at the Khayelitsha Cervical Cancer Screening Program (KCCSP) in 2003-4 were collected from study databases and CHW logbooks. Daily weather data from the South African Weather Service were linked for each appointment date. Cost data, presented in 2004 Rand, were collected from KCCSP's accounting system. For patients who did not attend as scheduled, we identified how many CHW visits they received and whether they ever returned. Total CHW visit costs were divided by all CHW visits performed to calculate the mean cost per visit. In the actual study, CHW visits were made until patients returned or refused. We evaluated the counterfactual of stopping after a given number of CHW visits, calculating the cost of CHW visits needed to achieve that particular return rate. We used multiple logistic regression to identify baseline predictors of spontaneous return. For those individuals who did not spontaneously return, we used multiple logistic regression to identify predictors of who was likely to return after CHW visits.
Results: Spontaneous return rates ranged from 52-75%, depending on appointment type. With 4-9 CHW visits, return rates were 76-92% and cost 13-41 Rand/patient. 69-89% return could be achieved at 8-13 Rand/patient (1 CHW visit), and 73-90% at 11-23 Rand/patient (2 CHW visits). Having lower socioeconomic status and having moved to Cape Town recently were each independently associated with lower return rates and with lower return after CHW visits.
Conclusions: The function of cost to increase return rate is useful for performing cost-effectiveness analyses to evaluate optimal levels of CHW effort allocated for this purpose. Subgroups identified as being less likely to return and less likely to respond to CHW interventions are prime candidates for preventive interventions since the cost and number of CHW visits needed are high.