Background: Patient-reported measures including health confidence correlate with health outcomes, such as hospital utilization and post-acute care (PAC) needs within research settings. A more recent approach has been to develop learning health systems (LHS) with these measures to incorporate research findings into clinical care at a rapid rate. We have found that a measure of health confidence, the health confidence score (HCS), collected by our research team, is significantly related to length of stay (LOS) and PAC needs, but not readmission. To explore the potential value of LHS approaches, we aimed to understand within a clinical pilot to see if the same patterns between the HCS and hospital utilization and PAC needs are observed when the HCS is collected through routine clinical processes.
Methods: On admission to a general medicine floor at a large academic medical center, social workers conduct initial assessments to characterize the socioeconomic, home living environment, and discharge needs of the patient. The HCS was inserted into the initial assessment template and asked on one general medicine unit from July 11 to August 30, 2024. Data was then combined with chart-reviewed data of the patient’s LOS, readmission at 30 and 90 days, and PAC disposition. Population characteristics were defined by frequencies or mean and standard deviation (SD). We used bivariate analysis with Chi-squared and ANOVA to compare prior research values, readmission, and disposition. Linear, logistic, and multinomial logistic regression were used to assess the relationship of HCS with LOS readmission, and PAC disposition respectively. Statistical significance was p< 0.05.
Results: A total of 50 patients were collected within the study period. The average HCS was 7.24 (SD 3.86, 0-12), which was significantly lower (p< 0.01) than measured with a research approach. The average LOS was 8.08 days (SD 5.67, 3-30 days), 15 (30%) were readmitted within 30 days, and 21 (42%) were readmitted within 90 days. The LOS was not significantly different (p=0.4) but readmission was higher than in the research population (both p< 0.01) and patients discharged to significantly different rates of PAC than in the research population (p< 0.01). Patients with increasing HCS by 1 point were on average in the hospital 0.44 days less (p=0.03, 95%CI [-0.84, -0.04]). Patients with increasing HCS were less likely to be readmitted within 30 days (OR 0.84, p=0.044, 95%CI [0.72, 0.995]), however, the relationship was not seen within 90 days (p=0.56). There was a decrease in HCS seen with increasing levels of PAC needs relative to discharging home (Table 1). These relationships were like prior research findings, except 30-day readmission was not associated with HCS in the research population.
Conclusions: Both the LHS and the research approaches demonstrate that low health confidence was associated with hospital utilization and PAC needs. However, the LHS approach identified a much higher percentage of patients with low health confidence than the research approach and a relationship between 30 day readmission and HCS. This demonstrates the importance of measuring the HCS within this clinical population and the particular value of LHS approaches that collect data as a part of usual clinical care. We plan to continue to use the HCS to identify, triage, and improve hospitalization outcomes for patients while improving our overall learning health system efficiency.
