Smart Discharges to improve post-discharge mortality among children with suspected sepsis in Uganda: A prospective before-after study

Background: Post-discharge mortality among children following an acute illness in low-resource settings is high and demands urgent attention. We aimed to assess the impact of Smart Discharges, a mortality prevention risk-differentiated approach to peri-discharge care among children under five years admitted with suspected sepsis. We conducted a before-after two-phase study with staggered implementation at six hospitals in Uganda. During a baseline period (Phase-1), clinical prediction algorithms for post-discharge mortality were developed based on clinical and socio-demographic data collected at hospital admission. Outcomes, primarily mortality within six months of discharge, were compared against an interventional period (Phase-2). In the Smart Discharges intervention each family was given soap, a mosquito net, their risk category (low/medium/high/very high), counselling, and educational materials regardless of risk stratification, after which the intensity of recommended follow-up care was determined by age (0-6, 6-60 months) and predicted post-discharge mortality risk. Overall, 13,051 children were enrolled: Phase-1, n=6,955; Phase-2, n=6,096. Characteristics were similar between groups, including mean predicted post-discharge mortality risk (6.3% Phase-1 vs. 5.9% Phase-2). With Smart Discharges, 2,331/3,891 (59.9%) medium/high/very high-risk families attending all their scheduled visits. The observed post-discharge mortality rate was 439 (6.3%) in Phase-1 vs. 296 (4.9%) in Phase-2; adjusted hazard ratio 0.77 (95%CI 0.67 to 0.90) favouring the intervention. In Phase-1, 1,313 (18.9%) children were re-admitted to hospital vs. 1,024 (16.8%) in Phase-2. A simple approach of risk assessment paired with education and engagement through scheduled follow-up after discharge is an appropriate strategy to improve child survival in low-resource settings. Objective(s): To evaluate the impact of a risk-differentiated peri-discharge care program in children under five years of age admitted with suspected sepsis in Uganda, based on the implementation of clinical prediction algorithms for post-discharge mortality; the development of these algorithms has been previously described. Methods: All data were collected at the point of care using encrypted study tablets and these data were then uploaded to a Research Electronic Data Capture (REDCap) database hosted at the BC Children’s Hospital Research Institute (Vancouver, Canada). At admission, trained study nurses systematically collected data on clinical, social and demographic variables. Following discharge, field officers contacted caregivers at 2 and 4 months by phone, and in-person at 6 months, to determine vital status, post-discharge health-seeking, and readmission details. Verbal autopsies were conducted for children who had died following discharge. For this analysis, data from both cohorts (0-6 months and 6-60 months) were combined and analysed as a single dataset. Z-scores were calculated using height and weight. Hematocrit was converted to hemoglobin. BCS score was created by summing all individual components. Analyses were conducted in R version 4.1.3 (R Foundation for Statistical Computing, Vienna, Austria), and RStudio version 2022.2.3 (RStudio, Boston, MA). Limitations: This study was quasi-experimental and did not randomize either individuals or clusters, such as hospitals. Although the children in the two phases showed similar characteristics, potentially clinically significant differences included higher proportions of children with hypoxemia and severe malnutrition in Phase-1 and higher proportions of children with malaria, severe anemia, and hypoglycemia in Phase-2, which may have impacted survival rates. Consequently, we cannot be confident that temporal confounding did not play a role in the differential results between phases. Several multivariable analyses were conducted, all of which demonstrated similar pooled effects, lending confidence to our interpretation that our estimated interventional effect approximated the true effect. Secondly, the interventional approach encompassed several concurrent interventions. It is not possible to deduce the components that contributed most to the observed impact. Adherence to follow-up visits was good, but additional strategies to provide a more robust and effective linkage to follow-up care in the community may further improve the intervention’s effectiveness. Thirdly, the study’s inclusion criteria restricted participants to those with a proven or suspected infectious illness, which does not fully reflect a census sample of all admissions but is most likely to be the target of a potential scaling intervention. Finally, the contribution of risk prediction itself to the impact of the intervention could not be determined. The knowledge of an objective risk classification could have profound effects on the behavior of clinicians and caregivers, beyond those due to the other aspects of the intervention. With the relative infrequency of risk-differentiated care, it is difficult to know how this approach would work in the absence of formal risk calculations. This has implications for regions where the validity of post-discharge risk prediction is questionable and suggests a need to further develop and validate risk prediction models more widely. Funding source(s): The study was supported by funding from Grand Challenges Canada (#TTS-1809-1939), Thrasher Research Fund (#13878), BC Children’s Hospital Foundation, and Mining4Life.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.