Document Type

Capstone Experience

Graduation Date

5-2019

Degree Name

Master of Public Health

Department

Biostatistics

First Committee Member

Raymond Neff

Second Committee Member

Christopher Wichman

Third Committee Member

Jiangtao Luo

Fourth Committee Member

Fabio Almeida

Abstract

Abstract

Objective: to identify Health Program improvement opportunities including reducing risk of unplanned visits.

Background: Health Program is a program in a Midwest community assisting persons with diabetes and Medicaid to help patients better self-manage their diabetes and connect with community resources. Communities are challenged with helping a growing number of persons with diabetes with multiple chronic conditions, medications, and increased use of unplanned visits. Information was needed to understand unmet needs of this population associated with unplanned visits to develop program improvements.

Methods: In a quality evaluation of a single center, a model of baseline prognostic factors associated with risk of unplanned events was developed. A total of 98 cases enrolled 2011-2017 with PH Medicaid insurance had adequate data available for modeling diabetic patients. Descriptive analysis was conducted comparing two PCP groups (SH, non-SH): T-tests were used to compare means, Chi-square tests for proportions. Survival analysis was conducted using multivariable Cox regression with stepwise selection to derive a predictive model of unplanned events. Schoenfeld residuals tests ensured the Cox proportionality assumption was met. A modified bootstrapping method was used to validate the model.

Results: A total of 41 patients had unplanned events and 57 were censored. PCP group was not a significant predictor of unplanned events; evidence of differences between PCP groups were only found in medications prescribed, and one chronic condition. Survival analysis provided evidence of significant association (P<.001) between freedom from unplanned visits and several predictors: diagnosed schizophrenia/bi-polar (HR 4.6, p=.0085), depression medication (HR 3.6, p=.0035), statins (HR 2.8, p=.0068), A1c<7.0 (HR .24, p=.0023), A1c>9.0 (HR .20, p=.0004), pain medication (HR .09, p=<.0001), COPD also included in the model for clinical value (HR .140, p=.0609). The highest risk patient groups were identified by distinct separation in survival curves for these covariate sets: (1) at highest risk: schizophrenia/bi-polar diagnosis with pain meds (none: depression meds, statins, COPD diagnosis), (2) schizophrenia diagnosis with no other meds or COPD, (3) depression and pain meds (no schizophrenia or COPD diagnosis or statins), (4) schizophrenia/bi-polar diagnosis with pain meds. Modified bootstrapping validated this model to have 68% sensitivity, and 86% specificity.

Conclusions: To improve outcomes for Health Program enrollees and reduce unplanned visits, initiatives that address unmet needs of the highest risk groups could be most successful for both the patient and the health care system. Suggestions include: (1) equipping the care team nurse to facilitate appointment scheduling with patients and mental health (MH) providers including med reconciliation by the mental health provider, and an option for in-home via telehealth MH sessions, (2) the model could be packaged into a tool for nurses to estimate risk for individuals perhaps an app, spreadsheet, or an auto-flag in EHR report for new referrals, (3) if preceptor IRB were to provide authorization for sharing some information in this evaluation with the scientific community and other community health systems, other communities could benefit toward progress in improving outcomes and reducing risk of unplanned visits for diabetic Medicaid populations.

Download

Share

COinS