Objective Globally, one third of deaths each year are from cardiovascular diseases, yet no strong evidence supports any specific method of CPR instruction in a resource-limited setting. We hypothesized that both existing and novel CPR training programs significantly impact skills of hospital-based healthcare providers (HCP) in Botswana. Methods HCP were prospectively randomized to 3 training groups: instructor led, limited instructor with manikin feedback, or self-directed learning. Data was collected prior to training, immediately after and at 3 and 6 months. Excellent CPR was prospectively defined as having at least 4 of 5 characteristics: depth, rate, release, no flow fraction, and no excessive ventilation. GEE was performed to account for within subject correlation. Results Of 214 HCP trained, 40% resuscitate ≥1/month, 28% had previous formal CPR training, and 65% required additional skills remediation to pass using AHA criteria. Excellent CPR skill acquisition was significant (infant: 32% vs. 71%, p < 0.01; adult 28% vs. 48%, p < 0.01). Infant CPR skill retention was significant at 3 (39% vs. 70%, p < 0.01) and 6 months (38% vs. 67%, p < 0.01), and adult CPR skills were retained to 3 months (34% vs. 51%, p = 0.02). On multivariable analysis, low cognitive score and need for skill remediation, but not instruction method, impacted CPR skill performance. Conclusions HCP in resource-limited settings resuscitate frequently, with little CPR training. Using existing training, HCP acquire and retain skills, yet often require remediation. Novel techniques with increased student: instructor ratio and feedback manikins were not different compared to traditional instruction.