The Impact of Healthcare Provider Collaborations on Patient Outcomes: A Social Network Analysis Approach
Care of patients with chronic conditions is complicated and usually includes large number of healthcare providers. Understanding the team structure and networks of healthcare providers help to make informed decisions for health policy makers and design of wellness programs by identifying the influencers in the network. This work presents a novel approach to assess the collaboration of healthcare providers involved in the care of patients with chronic conditions and the impact on patient outcomes.
In the first study, we assessed a patient population needs, preventive service utilization, and impact of an onsite clinic as an intervention on preventive service utilization patterns over a three-year period. Classification models were developed to identify groups of patients with similar characteristics and healthcare utilization. Logistic regression models identified patient factors that impacted their utilization of preventive health services in the onsite clinic vs. other providers. Females had higher utilizations compared to males. Type of insurance coverages, and presence of diabetes/hypertension were significant factors that impacted utilization. The first study framework helps to understand the patient population characteristics and role of specific providers (onsite clinic), however, it does not provide information about the teams of healthcare providers involved in the care process.
Considering the high prevalence of diabetes in the patient cohort of study 1, in the second study, we followed the patient cohort with diabetes from study 1 and extracted their healthcare providers over a two-year period. A framework based on the social network analysis was presented to assess the healthcare providers’ networks and teams involved in the care of diabetes. The relations between healthcare providers were generated based on the patient sharing relations identified from the claims data. A multi-scale community detection algorithm was used to identify groups of healthcare providers more closely working together. Centrality measures of the social network identified the influencers in the overall network and each community. Mail-order and retail pharmacies were identified as central providers in the overall network and majority of communities. This study presented metrics and approach for assessment of provider collaboration. To study how these collaborative relations impact the patients, in the last study, we presented a framework to assess impacts of healthcare provider collaboration on patient outcomes.
We focused on patients with diabetes, hypertension, and hyperlipidemia due to their similar healthcare needs and utilization. Similar to the second study, social network analysis and a multi-scale community detection algorithm were used to identify networks and communities of healthcare providers. We identified providers who were the majority source of care for patients over a three-year period. Regression models using generalized estimating equations were developed to assess the impact of majority source of care provider community-level centrality on patient outcomes. Higher connectedness (higher degree centrality) and higher access (higher closeness centrality) of the majority source of care provider were associated with reduced number of inpatient hospitalization and emergency department visits.
This research proposed a framework based on the social network analysis that provides metrics for assessment of care team relations using large-scale health data. These metrics help implementation experts to identify influencers in the network for better design of care intervention programs. The framework is also useful for health services researchers to assess impact of care teams’ relations on patient outcomes.