Selection of Prediction Method of Basic Statistical Work Parameters of N.V. Sklifosovsky Research Institute for Emergency Medicine of the Moscow Healthcare Department

Background The most important part of the state social and economic policy is optimization of the healthcare system, where the loss of public health leads to economic damage. Against this background, forecasting the work of medical institutions is the basis for the successful development of healthcare, despite the fact that the healthcare system, indicators and standards of medical and social welfare are still not stable, and a clear development strategy for the shortand long-term period has not been worked out. Aim of study Determining the most optimal method for predicting the work of a medical institution, based on identification of the main trends in the time series when constructing a model of the dependence of parameters or determining the behavior of data as a stochastic series (i.e. modeling random processes and random events with some random error).

Material and methods To predict the main statistical indicators of N.V. Sklifosovsky Research Institute for Emergency Medicine based on a retrospective analysis, data were used that were submitted to the City Bureau of Medical Statistics and entered into official reporting forms (form № 30, approved by Goskomstat of the Russian Federation dated September 10, 2002, № 175): the number of hospitalized patients and mortality rates in inpatient and intensive care units. To select the optimal methodology for the experimental forecast model, data were used for the period from 1991 to 2016. Indicators for 2017 were taken as control values.

Results As a result of the comparison of several methods (moving averages, least squares approach, Brown model, Holt–Winters method, autocorrelation model, Box–Jenkins method) as applied to the work of N.V. Sklifosovsky Research Institute for Emergency Medicine, the Holt–Winters model was chosen as the most appropriate one for the data characteristics.

Findings 1. When using methods of moving averages, least squares, Box-Jenkins, as well as Brown model and autocorrelation, the forecast result is not always influenced by strictly straight-line indicators of the time series, due to the heterogeneity of the time series and the presence of outliers (often found in a medical institution providing emergency care), which lead to a significant decrease in the reliability of forecasting. 2. The application of the Holt–Winters model, which takes into account the exponential trend (the trend of time series indicators) and additive season (periodic fluctuations observed in the time series), is most suitable for processing statistical data and forecasting for long-term, medium-term and short-term periods taking the specifics of a hospital providing emergency care into account. 3. The choice of the optimal method for predicting the work of a medical institution, based on the identification of the main trends in the time series, taking most of the features in the modeling of random processes and events into account, allowed to reduce the relative forecast error.

1. Karasev NA, Ermolov AS, Turko AP, Kurilin BL, Kislukhina EV. Vliyanie reanimatsionnoy obespechennosti na rezul’taty lecheniya ostroy khirurgicheskoy patologii organov bryushnoy polosti v mnogoprofil’nykh bol’nitsakh g. Moskvy. Moskovskiy khirurgicheskiy zhurnal. 2012; (1): 48–54. (in Russ.)

2. Khubutiya MSh, Karasev NA, Kurilin BL, Kislukhina EV, KiselevskayaBabinina IV, Molodov VA. Razvitie reanimatsionnogo koechnogo fonda v mnogoprofil’nykh statsionarakh g. Moskvy i ego vliyanie na rezul’taty lechebnoy deyatel’nosti. Skoraya meditsinskaya pomoshch’. 2012; 13(3): 45–50. (in Russ.)

3. Yermolov AS, Smirnov SV, Karasev NA, Kurilin BL, Kislukhina EV, Kiselevskaya-Babinina IV, et al. The Analysis of the Main Work Data in the Moscow City Burn Center after Remodeling. Russian Sklifosovsky Journal “Emergency Medical Care”. 2016; (1): 60–62. (in Russ.)

4. Khubutiya MSh, Karasev NA, Kislukhina EV, Vasilyev VA, Kurilin BL, Medvedeva AB, et al. Analysis of Clinical and Organizational Activities in the N.V.Sklifosovsky Research Institute for Emergency Medicine in 2005–2015. Russian Sklifosovsky Journal “Emergency Medical Care”. 2016; (2): 59–63. (in Russ.)

5. Lukashin YuP. Adaptivnye metody kratkosrochnogo prognozirovaniya vremennyh ryadov. Moscow: Finansy i statistika Publ.; 2003. (in Russ.)

6. Svetun’kov SG. Metody social’no-ehkonomicheskogo prognozirovaniya: in 2 vol. Saint Peterburg: Izd-vo SpbGUEH Publ.; 2009. (in Russ.)

7. Linnik YuV. Metod naimen’shikh kvadratov i osnovy matematikostatisticheskoy teorii obrabotki nablyudeniy. Moscow: Fizmatgiz Publ.; 1958. (in Russ.)

8. Pokrovskiy VI, Briko NI (ed.) Obshchaya epidemiologiya s osnovami dokazatel’noy meditsiny. 2-e izd., ispr. i dop. Moskva: GEOTAR-Media; 2012. 2nd, rev. and ext. Moscow: GEOTAR-Media Publ.; 2012. (in Russ.)

9. Savilov ED, Mamontova LM, Astaf’ev VA, Zhdanova SN. Primenenie statisticheskih metodov v ehpidemiologicheskom analize. 2-e izd., dop. i pererab. Moscow: MEDpress-inform; 2004. (in Russ.)

10. Winters PR. Forecasting sales by exponentially weighted moving averages. Management Science. 1960; 6(3): 324–342.

11. Orlov AI. Ekonometrika. Moscow: Feniks Publ.; 2009. (in Russ.)

12. Pogodin SK. Metody otsenki portfeley investitsiy, vklyuchayushchikh tsennye bumagi i nedvizhimost’. Cand. econ. sci. diss. Moscow, 2006. (in Russ.) Available at: https://dlib.rsl.ru/viewer/01003277680#?page=1 [Accessed August 12, 2019].

13. Urbakh VYu. Matematicheskaya statistika dlya biologov i medikov. Moscow: Izd-vo Akademii nauk SSSR Publ.; 1963. (in Russ.)

14. Gubinova TV. Primenenie eksponentsial’nogo sglazhivaniya dlya prognozirovaniya vyruchki. Materialy VII Mezhdunar. stud. nauch. foruma–2015. Available at: https://scienceforum.ru/2015/article/2015008484 (in Russ.)

15. Osipov LA, Krichevsky AM. Estimation and application of time series models with long memory in economical problems. Informatsionnoupravliaiushchie sistemy. 2007; (5): 45–51. (in Russ.)

16. Makridakis S, Andersen A, Carbone R, Fildes R, Hibon M, Lewandowski R, et al. The accuracy of extrapolation (time series) methods: Results of a forecasting competition. J Forecast. 1982; 1(2): 111–153.