Publication
PREDICTING THE NEED FOR PALLIATIVE AND HOSPICE CARE FOR CHILDREN IN UKRAINE USING THE MOVING AVERAGE TREND METHOD
Nesterenko V, Shevchenko A, Shevchenko V, Riga O, Redka I · Neonatology, Surgery and Perinatal Medicine · 2025
Palliative and hospice care (PHC) is provided to patients with life-limiting conditions during the terminal phase of illness, typically in the final weeks of life. Globally, an estimated 56.8 million individuals require such care annually, of whom 25.7 million are in the last year of life. The prevalence of individuals in need of PHC remains stable at approximately 1% of the general population and exhibits a gradual annual increase. A substantial proportion of these patients are children. In Ukraine, estimates of the paediatric and adult population requiring palliative care range from 130000 to 2 million; however, there is no officially endorsed national list of conditions designated for state-funded palliative care provision. In 2018, the Ukrainian Centre for Social Data conducted an assessment of PHC needs and proposed inclusion of the following conditions: congenital malformations, severe perinatal disorders, cerebral palsy, malignant neoplasms, diabetes mellitus, severe and profound intellectual disability, HIV/AIDS, inflammatory diseases of the central nervous system, decompensated cardiovascular diseases, tuberculosis, phenylketonuria, cystic fibrosis, chronic hepatitis, and mucopolysaccharidoses. The estimated need was approximately 66000 individuals. During 2019 and 2020, this estimate declined, conforming to a downward linear trend. However, this method proved insufficiently accurate for forecasting future requirements. Aim of the study to identify the optimal method for predicting the need for palliative and hospice care among children with incurable conditions in Ukraine, with estimation of annual demand during the pre-war period. Materials and methods of research. The selection of a predictive modelling approach for refinement was conducted among linear and nonlinear trends (logarithmic and exponential). Although nonlinear models offer the advantage of non-negative outputs (avoiding physically implausible negative values), the linear trend demonstrated superior empirical fit and was therefore selected for enhancement. It was subsequently integrated into a time-series framework using the Seasonal Autoregressive Integrated Moving Average (SARIMA) model, with model selection criteria set at a 5% significance level for the Bukovyna State Medical University hypothesis and a 95% confidence interval (CI). The 2020 pre-war dataset was adopted as the reference model. Results. The final forecasting model corresponded to a moving average trend with a fixed smoothing window. The need for PHC among children in Ukraine in 2020, calculated using official national medical statistics, amounted to 49000 individuals – a reduction of 7.8% compared with 2019. The linear trend forecasted a decrease of 9.2%, whereas the moving average trend predicted a decline of 11.4%. The largest absolute discrepancies between forecasted and observed values were observed for decompensated cardiovascular diseases (–26.3%), malignant neoplasms (+11.5%), and intellectual disability (–10.7%). The most accurate prediction was obtained for tuberculosis. The moving average trend exhibited greater overall accuracy (mean absolute percentage error: 3.6%) compared with the linear trend (8.2%), albeit with marginally lower statistical significance. Evaluation of the 95% CI indicated high precision for conditions with stable epidemiological patterns. The hybrid model achieved statistical significance (p0.05) for 6 out of 14 nosological entities, compared with 4 out of 14 for the linear trend alone. Validation of the refined moving average forecasting method demonstrated that 7 out of 14 nosologies (50%) fell within the 95% CI. For 9 out of 14 conditions, the moving average method yielded a smaller prediction error. Overall, the moving average trend outperformed the linear trend in 64.0% of cases. The hybrid approach – combining linear trend extrapolation with moving average smoothing – enabled narrowing of the confidence interval and attainment of p 0.05 for 6 out of 14 nosologies, versus 4 out of 14 for the unmodified linear model. Conclusion . A hybrid moving average trend method with a constant smoothing window – derived from a linear trend refined through time-series analysis at the 5% Bukovyna State Medical University hypothesis threshold and 95% CI – enhanced the accuracy of predicting paediatric palliative and hospice care needs in Ukraine. Nevertheless, model calibration remains necessary for critical nosologies, such as decompensated cardiovascular diseases at palliative stages. Utilisation of the confidence interval for the moving average trend facilitated identification of both model strengths and priority areas for methodological improvement © 2025, Bukovyna State Medical University. All rights reserved.
Synced from the TRENDS in Pediatric Palliative Care Zotero library, curated by The Siden Research Team. ICPCN does not host or verify the full text.