Abstract
Objectives
Existing predictors of psychiatric hospitalization are often limited to specific diagnoses or clinical scenarios. This study proposed an operationalization of clinical instability based on individual-level in Clinical Global Impression-Severity (CGI-S) scores and investigated its association with risk of psychiatric hospitalization.
Methods
This retrospective cohort study used data from 36,914 patients with seven psychiatric diagnoses from NeuroBlu, a real-world data repository containing de-identified electronic health record data from US mental healthcare providers. Patients were required to have ≥5 CGI-S measures during a 2-month window to ensure robust estimates of clinical severity and instability. Clinical severity was captured by the mean CGI-S. Clinical instability was a time-adjusted root mean square successive differences (RMSSD) of the CGI-S. The primary outcome was any inpatient visit within 6 months after the end of the 2-month window. Time-to-event analysis using Cox proportional hazard models adjusted for demographic and clinical characteristics was used to calculate hazard ratios (HR). Statistical significance was set at two-tailed p-value
Results
Clinical instability and severity independently predicted the risk of hospitalization (HR = 1.09; 95% confidence interval [CI] 1.07–1.10 and 1.11; 95% CI 1.09–1.12, for every standard deviation increase in instability and severity respectively, p
Conclusions
This study shows that clinical instability may serve as a transdiagnostic predictor of psychiatric hospitalization, providing information complementary to clinical severity. This development could improve outcome prediction in real-world clinical practice.
Existing predictors of psychiatric hospitalization are often limited to specific diagnoses or clinical scenarios. This study proposed an operationalization of clinical instability based on individual-level in Clinical Global Impression-Severity (CGI-S) scores and investigated its association with risk of psychiatric hospitalization.
Methods
This retrospective cohort study used data from 36,914 patients with seven psychiatric diagnoses from NeuroBlu, a real-world data repository containing de-identified electronic health record data from US mental healthcare providers. Patients were required to have ≥5 CGI-S measures during a 2-month window to ensure robust estimates of clinical severity and instability. Clinical severity was captured by the mean CGI-S. Clinical instability was a time-adjusted root mean square successive differences (RMSSD) of the CGI-S. The primary outcome was any inpatient visit within 6 months after the end of the 2-month window. Time-to-event analysis using Cox proportional hazard models adjusted for demographic and clinical characteristics was used to calculate hazard ratios (HR). Statistical significance was set at two-tailed p-value
Results
Clinical instability and severity independently predicted the risk of hospitalization (HR = 1.09; 95% confidence interval [CI] 1.07–1.10 and 1.11; 95% CI 1.09–1.12, for every standard deviation increase in instability and severity respectively, p
Conclusions
This study shows that clinical instability may serve as a transdiagnostic predictor of psychiatric hospitalization, providing information complementary to clinical severity. This development could improve outcome prediction in real-world clinical practice.
| Original language | English |
|---|---|
| Article number | CO199 |
| Pages (from-to) | S52 |
| Number of pages | 1 |
| Journal | Value in Health |
| Volume | 26 |
| Issue number | 6 |
| DOIs | |
| Publication status | Published - 1 Jun 2023 |