| Objective: A ten years follow-up study was carried out to understand survivalconditions of stroke patients and analyze the factors associated with survival and bothCox-regression model and parametric method was applied to estimate the survival andto identify significant prognostic factors. The difference between two mathematicalmethods was analyzed.Methods: From Nov. 1995 to Dec. 1996, 189 stroke patients were selected in theDepart -ment of Neurology, the General Hospital, Tianjin Medical University. Allcases were diagnosed according to cerebrovascular diagnostic criteria revised byChinese Medical Association Second National Conference on Cerebrovasculardiseases. All patients were diagnosed within 48 hours after stroke attack with CTscan. The patient must have lived in Tianjin for more than 15 years. The cases werefollowed up in 3 months, 6 months, 3 years, 7 years, 9 years and 10 years since theenrollment of the study. The data collected included the history of illness, physicalexamination, laboratory examination, and survival conditions. The work of follow-upended on Dec. 2006. For the statistical analyses, t test and X2 test were used fordescription of data, Life-Table and Kaplan-Meier methods were used for survivaldescription, and Log- Rank test was used for the significance test of effect ofprognostic factors. Cox regression and Weibull parametric model were used as mainstatistical methods for prognostic factors analyses. Bootstrap resampling method wasused for parameter estimation and hypothesis test of regression coefficient.Results: 1. At the end of follow-up, Of 189 patients, 105 patients were dead. Maincauses of death were stroke and cardiovascular disease. The Survival rate was 79.37%in one year, 64.02% in three years, 57.46% in seven years and 53.89% in ten years.2. Factors associated with survival of stroke by Cox regression: Age (RR=1.035,P＜0.05), work type(RR=1.507,P＜0.05), physical exercises before stroke(RR=0.323,P＜0.001), history of heart disease (RR=2.044, P＜0.05), hypertension history (RR=1.707, P＜0.05) and stroke history(RR=1.856, P＜0.05), the factors associated with severity of illness such as body temperature, consciousness when came to thehospital and nervous system symptom; Brain injury condition such as area of brainimpact (RR=1.023, P＜0.001), existence of brain hernia, agnosia and anepia.Laboratory examinations such as blood glucose level (RR=1.006, P＜0.001); and theconditions when discharge from hospital, rehabilitative treatment of post-dischargeand recurrence are also statistically significant. Social psychic factors such asconfidence of rehabilitation, repression, negative life event, support from relatives andfriends were also related to prognosis of stroke.3. Results by Parametric method: (1) With tests of goodness of fit and distributionselection, Weibull distribution is optimal (Log Likehood Ratio= 107.838384, P＜0.05;RNL=0.9642, P＞0.05). (2) Result of Weibull regression model: The main strokesurvival factors with statistical significance include: age (RR=1.085,P＜0.05), physicalexercises before stroke (RR=0.048,P＜0.001), cardiac disease history (RR=9.885,P＜0.05), hypertension history (RR=8.365, P＜0.05) and stroke history (RR=2.852,P＜0.05), the factors associated with severity of illness such as body temperature,consciousness when came to the hospital and nervous system symptom. Brain injurycondition such as area of brain impact (RR=1.036, P＜0.001), brain hernia, agnosiaand anepia. Laboratory examinations such as blood glucose level (RR=1.019,P＜0.001); and the conditions when discharge from hospital and rehabilitative treatment of post-discharge are also statistically significant. Furthermore, we found thatsocial psychic factor such as rehabilitative confidence, repression, negative event,support from relatives and friends were also related to prognosis of stroke. Contrastwith the results of Cox regression, recurrence is not included in the main strokesurvival factors with statistical significance.4. Bootstrap resampling method: The Regression coefficient' confidence interval ofbootstrap is much narrower than some of Cox regression. Meanwhile, the results ofhypothesis test of the effects of prognostic factors are different from some of Coxregression.Conclusions: Both Cox and parametric method in analyzing the present survival data are effective. Hypertension history, stroke history, cardiac disease history andbrain injury condition and nervous system symptom are related to survival of stroke.The patients persisting physical exercises before stroke tend to have the bestprognosis. Some Social psychic factors also are related to the prognosis. |
PDF Full Text Request