| Sažetak | Cilj istraživanja: Cilj istraživanja bio je procijeniti kvalitetu intenzivnog liječenja (KIL) u Jedinici intenzivnog liječenja I, Klinike za anesteziologiju i intenzivno liječenje Kliničkog bolničkog centra Rijeka (JIL I) Simplified Acute Physiology Score II (SAPS II) sustavom. Također, cilj ovog istraživanja bio je utvrditi prediktivnu vrijednost nagiba (koeficijenta smjera) regresijskog pravca (kn) i nove vjerojatnosti smrtnog ishoda (VSI) intenzivnog liječenja novim SAPS II sustavom (noviPrSIIn) izračunatih iz niza VSI SAPS II sustava (PrSIIn) po danima (n) intenzivnog liječenja.
Ispitanici i metode: Prospektivno su obrađeni podaci o 922 bolesnika liječena u JILu I tijekom jednogodišnjeg razdoblja. Primjenom isključnih kriterija u istraživanju KILa preostalo je 654 bolesnika, a u istraživanju značaja kn 354 bolesnika. SAPS II vrijednosti u bodovima za sve bolesnike do petnaestog dana intenzivnog liječenja su upotrebom originalne formule SAPS II sustava pretvorene u PrSII1-15. Iz srednje vrijednosti svih PrSII1 i zamijećene bolničke smrtnosti izračunat je Standardizirani omjer bolničke smrtnosti (SOBS). Diskriminacija i kalibracija SAPS II sustava su procijenjene ROC (od engl. Receiver operating characteristic) krivuljom i Hosmer-Lemeshowljevih testovima. Za sve PrSII2-15 su jednadžbom pravca regresije izračunati k2-15 i noviPrSII3-15. PrSIIn i noviPrSIIn u izračunu, za razliku od PrSIIn, sadržavaju povijest svih dotadašnjih dana intenzivnog liječenja. Bolesnici su razvrstani prema: dobi (<40, 40-60 i >60 godina), spolu i načinu prijema (kirurški elektivni, KE; kirurški hitni, KH i medicinski, M) kao i prema PrSII1 u: ≥0%, ≥30% i ≥50% PrSII1. Preciznost kn, PrSIIn i noviPrSIIn je procijenjena mjerenjem i usporedbom površina ispod ROC krivulja u istih populacija bolesnika. Modelom logisitičke
regresije (LR), unatražnjim isključivanjem je utvrđeno koje varijable preostaju kao značajni prediktori ishoda intenzivnog liječenja (IIL) u sveukupne populacije i u podskupinama.
Rezultati: Demografska obilježja bolesnika su bila slična populacijama u istraživanjima SAPS II sustava iz zapadnih zemalja (prosječna dob 62 godine i 63%tna zastupljenost muškaraca) osim niskog učešća M bolesnika (24%) što je bio i razlog za niski stupanj bolesti pri prijemu (PrSII1=25%). Uz zabilježenu bolnička smrtnost od 23,7%, SOBS (0,95, 95% CI 0,805-1,101) je pokazao dobru KIL u JILu I. Diskriminacijska obilježja SAPS II sustava su dobra (ROC=0,835, 95% CI 0,805-0,863; P<0,001). SAPS II sustav dobro odjeljuje bolesnike koji neće preživjeti intenzivno liječenje. Ovaj sustav značajno precjenjuje zamijećenu predviđenom smrtnosti (Hosmer Lemeshow H, chi-kvadrat=28,98; df=8; P<0,001 i C test chi-kvadrat=11,27; df=8; P=0,020) što pokazuje lošu kalibriraciju. Veći odmaci dobrog predviđanja su pronađeni u skupina bolesnika s PrSII1≥80%. Demografska obilježja populacije u istraživanju kn i noviPrSIIn, osim veće učestalosti KH bolesnika (48,6%) su bila slična populaciji u istraživanju KIL. Najbolja diskriminacijska obilježja je pokazala PrSIIn, a vrlo slična noviPrSIIn varijabla (svi ROC≥0,700) te obje predstavljaju jednakovrijedne prediktore IIL. Površine ispod ROC krivulja za kn u niti jednom danu ne prelaze 0,600 pa se kn ne može preporučiti kao prediktor IIL. U podskupina pronađena su bolja diskriminacijska obilježja kn i noviPrSIIn u odnosu na PrSIIn u bolesnika starije dobi i KE bolesnika. U bolesnika s PrSII1≥30% noviPrSIIn je pokazao bolja diskriminacijska obilježja u odnosu na PrSIIn. Statistički značajni prediktori IIL pronađeni u ovom istraživanju su u: sveukupne populacije PrSII3,4,5-7,14-15, k12-15 i noviPrSII4,5,15; dobne populacije od
40-60 godina: noviPrSII3, a >60 godina noviPrSII5,7; muškaraca: k4 i noviPrSII4; KE bolesnika: noviPrSII3,5,7; bolesnika s PrSII1≥30%: PrSII3 i noviPrSII4,5,6; a s PrSII1≥50%: PrSII3 i noviPrSII4. Nažalost, u podskupinama je bio premali broj bolesnika za sigurne zaključke. Za potvrdu rezultata ovog istraživanja kao i pronalaženje izvora razlika u diskriminacijskim obilježjima predikcijskih varijabli IIL potrebno je višeintitucijsko istraživanje s datotekom od nekoliko desetaka tisuća bolesnika.
Zaključci: KIL u JILu I je u razini KIL zapadnog svijeta mjereno SOBSom iz SAPS II sustava koji je pokazao dobra diskriminacijska i loša kalibracijska obilježja. Nagibi regresijskih pravaca (kn) izračunati iz PrSIIn pokazuju lošija diskriminacijska obilježja u odnosu na PrSIIn te nisu preporučljivi kao prediktor IIL u sveukupne populacije liječene u JILu I. VSI noviPrSIIn pokazuju vrlo slična diskriminacijska obilježja u odnosu na PrSIIn te se obje varijable mogu koristiti kao prediktori IIL u sveukupne populacije u JILu I. U istrživanju je pronađeno više novih statistički značajnih prediktora IIL u sveukupne populacije i podskupina koji u budućnosti mogu postati dio novih sustava predviđanja IIL. |
| Sažetak (engleski) | Objectives: Aim of this study was to estimate quality of intensive care medicine (QICM) in Intensive Care Unit I, Department of Anaesthesiology and Intensive care medicine, University Hospital Rijeka (ICU I) by Simplified Acute Physiology Score II (SAPS II). Likewise, aim was to determine predictive value of regression trend (coefficient of the direction) (kn) and new predictor of mortality by new SAPS II (noviPrSIIn) for all days (n) during stay in ICU I for all patients.
Patients and Methods: During one year period, data from all of 922 patients treated in the ICU I prospectively was observed. Applying original exclusion criteria, 654 patients were included in study of QICM and 354 patients were included in the study of the kn. For all patients and for all days of ICU stay, 15th ICU day inclusive, SAPS II point’s value was converted to PrSIIn by original SAPS II formula. Mean of all PrSII1 was divided by observed hospital mortality and Standardized mortality ratio (SMR) was calculated. Discrimination and calibration of SAPS II was estimated by Receiver operating characteristic curve (ROC) and Hosmer-Lemeshow goodness-of-fit tests. For PrSII2-15, k2-15 and noviPrSII3-15 was calculated by regression direction formula. Both, kn and noviPrSIIn included, unlike PrSIIn, history of all of the previous ICU days. Patients were divided according to age: (<40, 40-60 and >60 years), sex and type of admission (elective surgical, emergency surgical and medical) as well as PrSII1 to: ≥0%, ≥30% and ≥50% PrSII1. Validation of the kn, PrSIIn and noviPrSIIn was estimated by measuring the area under ROC curves as well as by comparison of the areas in the same populations of the patients. In whole population as well as in the subgroups of the patients statistical significant variables were determined using backward option of the logistic regression.
Results: Demographic characteristics of the patients were very similar to characteristics of the patients from the SAPS II studies from western countries (average age was 62 years and there were 63% males). Low participation of the medical patients was exception what was a reason for low predicted mortality after ICU admission (PrSII1=25%). Observed hospital mortality was 23.7% and SMR (0.95, 95% CI 0.805-1.101) indicating a good QICM in the ICU I. Discrimination of the SAPS II was good (ROC=0.835, 95% CI 0.805-0.863; P<0.001). SAPS II has good properties to separate patients who do not survive intensive care treatment (ICT). This system significantly overestimates observed by predicted mortality (Hosmer Lemeshow H, chi-kvadrat=28.98; df=8; P<0.001 and C test chi-kvadrat=11.27; df=8; P=0.020) which indicate poor calibration of the SAPS II in the ICU I. Larger variances between obseved and predicted mortality were noted in the PrSII1≥80% patients. Demografic characteristic of the patients in the study of the kn and noviPrSIIn, except larger proportion of the emergency patients (48.6%) were similar to the patients in the study of the QICM. PrSIIn, and very similar noviPrSIIn, showed the best discrimination (ROC≥0.700) so both variables were presented equivalent good predictive value of the ICT. Areas under ROC curves for the kn weren't preponderate 0.600 so kn can't be proposed as a predictor of the ICT. In the subgroups, kn i noviPrSIIn, comparing with PrSIIn, showed the best discriminative properties in the older (>60 years) and elective patients. NoviPrSIIn showed the best discriminative properties in the subgroup of the patients with the PrSII1≥30%. Statisticaly significant predictors of the ICT by the varables and ICU I days of the patients stay founded in this study are in: whole study population: PrSII3,4,5-7,14-15, k12-15 and noviPrSII4,5,15; population between 40-60 years: noviPrSII3 and >60 years noviPrSII5,7; males: k4 and noviPrSII4; elective patients: noviPrSII3,5,7; patient with PrSII1≥30%: PrSII3 and noviPrSII4,5,6; and patients with PrSII1≥50%: PrSII3 and noviPrSII4. Unfortunately, number of the patients in the subgroups was to small for strong conclusions. To confirm the results of this study it is necessary to perform several ten thausands patients multicenter study.
Conclusions: QICM in an ICU I is comparable to QICM founded in the studies from the western countries measured by the SMR from SAPS II. This system showed good discrimination and poor calibration. Regression trend (kn) calculated from PrSIIn showed poorer dicrimination comparing with the PrSIIn and it can’t be recommended as a predictor of the ICT in whole population of the ICU I. The noviPrSIIn showed similar discrimination comparing with PrSIIn so both can be recomended as a predictors of the ICT in whole population of the ICU I. Several new statistically significant predictors of the ICT in the whole population as well as in the subgroups were founded in this study and all of these can be used in the future as a part of new prediction systems. |
