基于马尔可夫模型的重症社区获得性肺炎患者生理指标动态变化与短期预后OA

Objective:Severe community-acquired pneumonia(SCAP)is a common and critical infectious disease characterized by high incidence,rapid progression,and high mortality.Patients with SCAP often exhibit complex and dynamic disease progression during the early stage of hospitalization,and some patients may deteriorate rapidly or even experience sudden death.Identifying trends in disease progression and predicting short-term clinical outcomes through continuously monitored physiological indicators is an important research direction in critical care medicine.Traditional static statistical models are difficult to reflect the dynamic evolution of diseases,whereas multistate models can describe the transition processes of individuals between different health states,thereby more realistically reflecting disease progression patterns.The discrete-time Markov model(DTMM)can be used to analyze stage-wise transition probabilities between different clinical states,while the continuous-time Markov model(CTMM)can quantify the effects of covariates on the instantaneous risk of state transitions.This study aimed to construct DTMM and CTMM based on changes in key physiological indicators during the first 3 consecutive days of hospitalization in patients with SCAP,quantitatively evaluate the impact of dynamic physiological indicators on short-term clinical state transitions and mortality risk,and provide a statistical basis for early risk assessment and clinical intervention in patients with SCAP. Methods:This was a single-center retrospective study.Clinical data of patients with SCAP who were hospitalized at the People's Hospital of Boluo County between January 2024 and December 2024 were collected.Patients who met the diagnostic criteria for SCAP and had continuous physiological indicator records for three consecutive days were included.Baseline demographic characteristics,comorbidities,and daily physiological indicators were collected,including prealbumin(PAB),blood glucose(GLU),and blood urea nitrogen(BUN).Daily monitoring indicators were treated as time-varying covariates.According to daily disease progression,patients were categorized into stable state(State 1),deterioration state(State 2),and death state(State 3).First,a DTMM was constructed to analyze the transition probabilities between different clinical states,and the cumulative transition probabilities at 1 day and 3 days were calculated.Subsequently,multinomial logistic regression was used to evaluate the effects of covariates on transition probabilities.Furthermore,a CTMM was established by incorporating major physiological indicators into the model to quantify their effects on the instantaneous risk of state transitions,and hazard ratios(HRs)with 95%confidence intervals were calculated. Results:A total of 80 patients with SCAP were included.DTMM results showed that clinical state transitions during the first 3 days of hospitalization exhibited clear dynamic changes.The baseline transition probability matrix suggested a certain dynamic equilibrium among different states.Patients in State 2 still had some potential for recovery in the short term:In the PAB model,the cumulative probability of State 2 patients recovering to State 1 within 3 days was 64.1%;in the GLU model,it was 49.0%.Although some patients showed a recovery trend,the model also indicated a relatively high short-term risk of death.In the PCT model,the cumulative risk of death within 3 days was 20.5%for patients in State 1 and 19.5%for those in State 2.In both short-term(t=1 day)and medium-term(t=3 days)predictions,the risk of direct transition from Stable 1 to State 3 was similar to,or even slightly higher than,that from State 2 to State 3,suggesting that some SCAP patients may experience sudden death.CTMM results showed that PAB levels were negatively associated with mortality risk.For each one-unit increase in PAB,the instantaneous risk of direct transition from State 1 to State 3 decreased by approximately 2%(HR<1).In contrast,for each one-unit increase in GLU,the instantaneous risk of transition from State 1 to State 2 increased by approximately 29%(HR>1),indicating that elevated GLU is an important risk factor for patients transitioning from State 1 to State 3.Although BUN did not reach statistical significance,its point estimate of the hazard ratio(HR=2.46)suggested a potential risk trend. Conclusion:Analysis based on the multistate Markov model indicated that GLU and PAB are important predictors affecting the short-term outcomes of patients with SCAP.Higher PAB levels may have a protective effect against sudden death,whereas elevated GLU significantly increases the risk of transition from a stable state to a deteriorating state.Early disease progression in patients with SCAP exhibits clear dynamic characteristics.Although some patients with deterioration still have a relatively high potential for recovery in the short term,the overall risk of death and unpredictable sudden death events remains high.Therefore,strengthening GLU control and nutritional support in the early management of patients with SCAP may help improve short-term prognosis.Meanwhile,multistate Markov models provide an effective statistical tool for studying the dynamic progression of critical illnesses and may offer references for clinical risk assessment and individualized treatment decision-making.