A Novel Model for Predicting Rehospitalization Risk Incorporating Physical Function, Cognitive Status, and Psychosocial Support Using Natural Language Processing.

BACKGROUND: With the increasing focus on reducing hospital readmissions in the United States, numerous readmissions risk prediction models have been proposed, mostly developed through analyses of structured data fields in electronic medical records and administrative databases. Three areas that may have an impact on readmission but are poorly captured using structured data sources are patients' physical function, cognitive status, and psychosocial environment and support. OBJECTIVE OF THE STUDY: The objective of the study was to build a discriminative model using information germane to these 3 areas to identify hospitalized patients' risk for 30-day all cause readmissions. RESEARCH DESIGN: We conducted clinician focus groups to identify language used in the clinical record regarding these 3 areas. We then created a dataset including 30,000 inpatients, 10,000 from each of 3 hospitals, and searched those records for the focus group-derived language using natural language processing. A 30-day readmission prediction model was developed on 75% of the dataset and validated on the other 25% and also on hospital specific subsets. RESULTS: Focus group language was aggregated into 35 variables. The final model had 16 variables, a validated C-statistic of 0.74, and was well calibrated. Subset validation of the model by hospital yielded C-statistics of 0.70-0.75. CONCLUSIONS: Deriving a 30-day readmission risk prediction model through identification of physical, cognitive, and psychosocial issues using natural language processing yielded a model that performs similarly to the better performing models previously published with the added advantage of being based on clinically relevant factors and also automated and scalable. Because of the clinical relevance of the variables in the model, future research may be able to test if targeting interventions to identified risks results in reductions in readmissions.

Can Text Messages Improve Attendance to Primary Care Appointments in Underserved Populations?

We explored whether text message (TM) reminders could be used at a community health center (CHC) to improve primary care appointment attendance in adult patients. Over six months, we allocated 8,425 appointments to intervention and 2,679 to control. The proportion of no-shows in the intervention was 18.0% vs. 19.8% in control (p = .106). Among intervention appointments, 1,431 did not have a cell phone, 4,955 did not respond to the consent TM, and 231 declined TMs. The proportion of no-shows for the 1,309 appointments who received TM was 13.7% compared with 20.2% in a matched control group (p = .001). However, of 81 surveyed patients who did not respond to the consent TM, 64 (93%) wished to receive TMs. In conclusion, patients who received TM demonstrated improved attendance to their appointments. TM might be an effective supplemental appointment reminder method in a subpopulation of CHC patients and it should be explored in future research.

Patient navigation based on predictive modeling decreases no-show rates in cancer care.

BACKGROUND: Patient adherence to appointments is key to improving outcomes in health care. "No-show" appointments contribute to suboptimal resource use. Patient navigation and telephone reminders have been shown to improve cancer care and adherence, particularly in disadvantaged populations, but may not be cost-effective if not targeted at the appropriate patients. METHODS: In 5 clinics within a large academic cancer center, patients who were considered to be likely (the top 20th percentile) to miss a scheduled appointment without contacting the clinic ahead of time ("no-shows") were identified using a predictive model and then randomized to an intervention versus a usual-care group. The intervention group received telephone calls from a bilingual patient navigator 7 days before and 1 day before the appointment. RESULTS: Over a 5-month period, of the 40,075 appointments scheduled, 4425 patient appointments were deemed to be at high risk of a "no-show" event. After the patient navigation intervention, the no-show rate in the intervention group was 10.2% (167 of 1631), compared with 17.5% in the control group (280 of 1603) (P<.001). Reaching a patient or family member was associated with a significantly lower no-show rate (5.9% and 3.0%, respectively; P<.001 and .006, respectively) compared with leaving a message (14.7%: P = .117) or no contact (no-show rate, 21.6%: P = .857). CONCLUSIONS: Telephone navigation targeted at those patients predicted to be at high risk of visit nonadherence was found to effectively and substantially improve patient adherence to cancer clinic appointments. Further studies are needed to determine the long-term impact on patient outcomes, but short-term gains in the optimization of resources can be recognized immediately.

Success of automated algorithmic scheduling in an outpatient setting.

OBJECTIVE: To determine if algorithmically generated double-booking recommendations could increase patient volume per clinical session without increasing the burden on physicians. STUDY DESIGN: A randomized controlled trial was conducted with 519 clinical sessions for 13 dermatologists from December 1, 2011, through March 31, 2012. METHODS: Sessions were randomly assigned to "Smart-Booking," an algorithm that generates double-booking recommendations using a missed appointment (no-shows + same-day cancella- tions) predictive model (c-statistic 0.71), or to a control arm where usual booking rules were applied. The primary outcomes were the average number and variance of arrived patients per session, after controlling by physician. In addition, physicians received a survey after each session to quantify how busy they felt during that session. RESULTS: 257 sessions were randomized to Smart-Booking and 262 sessions were randomized to control booking. Using a generalized multivariate linear model, the average number of arrived patients per session was higher in the Smart-Booking intervention arm than the control (15.7 vs 15.2, difference between groups 4.2; 95% CI, 0.08-0.75; P = .014).The variance was also higher in the intervention than control (3.72 vs 3.33, P = .38).The survey response rate was 92% and the physicians reported being similarly busy in each study arm. CONCLUSIONS: Algorithmically generated double-booking recommendations of dermatology clinical sessions using individual physician assumptions and predictive modeling can increase the number of arrived patients without overburdening physicians, and is likely scalable to other settings.

Development and implementation of a real-time 30-day readmission predictive model.

Hospitals are under great pressure to reduce readmissions of patients. Being able to reliably predict patients at increased risk for rehospitalization would allow for tailored interventions to be offered to them. This requires the creation of a functional predictive model specifically designed to support real-time clinical operations. A predictive model for readmissions within 30 days of discharge was developed using retrospective data from 45,924 MGH admissions between 2/1/2012 and 1/31/2013 only including factors that would be available by the day after admission. It was then validated prospectively in a real-time implementation for 3,074 MGH admissions between 10/1/2013 and 10/31/2013. The model developed retrospectively had an AUC of 0.705 with good calibration. The real-time implementation had an AUC of 0.671 although the model was overestimating readmission risk. A moderately discriminative real-time 30-day readmission predictive model can be developed and implemented in a large academic hospital.