Evaluating the impact of the National Health Insurance Fund oncology benefits package and a healthcare workers’ strike on time to cancer treatment initiation in Nairobi County, Kenya: An interrupted time series analysis

Introduction: In April 2015, Kenya introduced the National Health Insurance Oncology Benefits Package and its complementary reforms (oncology insurance scheme) to alleviate financial hardship among its members upon a cancer diagnosis. In this study, we hypothesised that the time it took to start treatment would have an impact on health outcomes: the longer patients waited the worse their outcomes would be. We did not have outcomes in the data but we could compute time to treatment initiation (TTI). While assessing the impact of the oncology insurance scheme on TTI, we encountered a substantial sudden increase in average TTI in June 2018 which we needed to explore. Methods: We conducted our analysis using R, a statistical computing software, for interrupted time series analysis (ITSA) on Nairobi Cancer Registry data to assess the impact of the introduction of the oncology insurance scheme on TTI in days among Nairobi County residents diagnosed with cancer. We calculated the monthly median TTI, resulting in 120 data points (one for each of the 120 months of the observation period - January 1st 2010 to December 31st 2019). Since the oncology insurance scheme was available to the entire Kenyan population, a suitable control group was unavailable. To address this, we used auto regressive integrated moving average (ARIMA) modelling to forecast an expected trend, allowing us to estimate both sudden and gradual changes during April 2015 and June 2018 (intervention months). Results: After cleaning the data, 7584 (35%) cases of the original 21,464 were left for analysis. Females were more than males at 57.8%. Approximately 65% of the cases with known stage at diagnosis were in stages III and IV. No statistically significant impact was associated with the introduction of oncology insurance scheme; an additional 9.06 days (95% CI: −8.7 to 26.8) and a gradual change of 0.88 days per month (95% CI: −0.11 to 1.88). However, a statistically significant sudden increase in monthly median TTI in June 2018 of 34.6 days (95% CI 15.4 to 53.8) and the gradual change of −1.6 days (95% CI −3.5 to 0.4) per month which was not statistically significant, were associated with a healthcare workers’ strike. We could not accurately analyse case trends from these data because the registry had not completed collating data for the later years (2015–2019). Conclusions: These results suggest that the oncology insurance scheme may not have reduced average TTI for the cancer patients as we had hypothesized. However, a healthcare workers’ strike (based on corroboration with findings from the 2018 Kenya Household Health Expenditure and Utilization Survey), increased the average TTI among these patients. Data science techniques and ITSAs using cancer registry data is a cost-effective method to answer important population-level research questions in resource-limited settings.