As healthcare providers transition from fee-for-service to value-based care models and continue to look for ways to be more efficient in their day-to-day business, they are discovering new and innovative analytic models to identify high-risk patients, deploy evidence-based medicine and reduce adverse effects and infections. Therefore, they have to assemble information generated across the overall health and administrative systems, including data imported from outside sources and services.
In order to assemble this information efficiently and correctly, Big Data is often presented as a potential solution. However, especially in the case of a healthcare system, this data analysis is quite complex. Nevertheless, healthcare in some cases are presenting a very impressive use of analytics. In addition to the traditional providers like Siemens, GE, Philips, EPIC and Intersystems, several high-tech providers like Apple, IBM, Microsoft, Oracle and Google have been increasing investment in this area by buying startups, which are accelerating the healthcare digital transformation.
Healthcare tends to lag behind other industries in terms of IT governance and maturity. Effective use of analytics is not something one can buy from a vendor like medical ultrasonography equipment. It's an organizational and cultural value that has to grow and mature within the health system. In addition, its complexity is magnified by the shift the analytics have to make from merely reporting data to predicting data.
Big Data is commonly defined based on 3 V’s: large Volumes of data, Variety of data formats and high Velocity of processing. The common notion is that bigger is better when it comes to data and analytics, which implies high complexity and high investment required. This is not always the case. MIT professor Michael Stonebraker presented in his research that 69 percent of corporate executives named greater data variety as the most important factor. Therefore, many firms see the largest opportunity in Big Data from the use of traditional data sources that have gone untapped in the past. These are data sets that have typically sat outside the purview of traditional data warehouses – the “long tail” data.
Most hospitals have already implemented an Enterprise Resource Planning system, a Computerized Provider Order Entry, a Laboratory Information System, a Picture Archiving and Communication System, a Radiology Information System, an Electronic Health Record, a Medical Mobile Dictation system, a Health Information Exchange and/or an Electronic Medical Health suite, among others. Initially, it seems that the data available from these systems’ databases would be very valuable. Additionally, there are external data (population, demography, georeference data, illness occurrences, payers’ data, etc.) that can be aggregated to improve the analytic process.
However, in most of the cases, each system was implemented without data integration in mind and in different sequencing with old or new standards. Not all information is recorded by the medical staff, just the minimum required by the systems. Several areas of data are in fact in free text format. Some tables like medicine prescription are not normalized or even loaded at the systems.
Therefore, the actual business readiness (staffing and maturity level), the business capabilities (stable processes in place) and the technical capabilities (available IT software and hardware infrastructure) should be assessed.
It is necessary to recognize that the hospital or clinic should be managed based on data and, moreover, its medical efforts can be improved significantly by meaningful use of available data. First, a list of priorities and problems should be developed. Design thinking or other methodology could be applied to define potential solutions for each problem. At this point, the organization will be able to evaluate what is needed in order to achieve its goals. A business plan should be developed and approved by the board of the organization, as this program will likely affect several departments within the hospital. The discussions derived from this approval process will be important to develop the strategic and managerial objectives of the program, share knowledge about the subject and engage the top and middle management in addition to key stakeholders like doctors, nurses and the administrative staff.
A BI system may be enough to use as a starting point. In one case, a large healthcare provider adopted a BI solution, which allowed users to identify previously unknown evidence and indicators, helping them to meet the requirements of each problem area they had identified. In addition, this contributed to maximizing the company's revenue. Among several indicators, it was able to be accessed quickly – both the production as the cost of healthcare provider network, as well as identification of behavioral differences (the outliers) with providers. It was a huge advance for them.
The hospital must consider developing some core internal skills before pursuing such a program, such as establishing a proper and small competence center. An IT program always involves a learning curve for all. At times, full IT outsourcing contracts fail due to the client lack of internal skills to run this kind of operation.
In general, the data should start small, but the data needs to be concise and address the strategic healthcare problem being measured, not just the data that's easiest to obtain. While data that comes easy may make the project go faster, the analytic results are likely to have only limited value, which can jeopardize the whole program.
Another approach should be to consider cloud computing instead of building upon the healthcare system’s IT infrastructure from the beginning, based on the benefits of the cost per demand, processing capacity and HIPAA compliance. Several companies are providing cloud-based analytics platforms with several APIs for easy usage and maintenance. Although it is not without cost, it is much easier to implement than modifying some existing infrastructure, and requires much less expense initially.
Once the strategic objectives are broken down into managerial objectives and goals, they can be part of each stage of the overall road map.
The implementation program should build an iterative process of understanding data and applying it to business goals, looking for the following dimensions: financial, operational, clinical and behavioral. For each step, one should identify available data and assess its quality level, gather the business needs and technical options, define the technical solution, and test the potential results, using feedback from the experience for the next iteration. The idea is to establish a data science practice.
Based on the resulting road map, a complete IT architecture can be drawn as a reference model. However, it’s recommended to dissect the technology part of the map into simpler components that can be built and tested, considering the learning curve of the business and IT staff, and also the partners, staff and/or technologies. In addition, some architecturally significant decisions should be made, such as: cloud versus on-premise (privacy of data, latency, volume, cost), which technical platform to use, how to connect your existing clinical and enterprise data warehouses to it and how to integrate with existing services.
Conceptually, an Enterprise Master Patient Index (EMPI) is a database that is used across a healthcare organization to maintain consistent, accurate and current demographic and essential medical data on the patients seen and managed within its various departments. The patient is assigned a unique identifier that is used to refer to this patient across the enterprise. Therefore, the patient record must be identified individually across all data sets. However, this is not true everywhere, and keeping the data quality high on all data sets is not a trivial matter.
Clinical data contains codes using different coding schemes like CPT, ICD9, ICD10 or NDC and also may contain non-standard coding schemes, requiring translation services in order to have a common dictionary for the data analysis.
Data sources from the healthcare systems provide flat files in HL7 format that are not immediately useful for data processing. This then requires de-serializing and exposing HL7 messages as pre-defined structures to the rest of the clinical processing pipeline. Data about a given population is normally split across multiple flat files that contain hundreds of thousands of rows. However, not all data is accessible for processing due to the complexity it introduces to existing Extract-Transform-Load processes adopted by the BI systems.
Traditionally, data visualization or reporting is developed based on a waterfall approach where an expert provides the requirements in the earlier phases of the process and verifies the final result at the end. In Big Data, the agile approach is much more appropriate, where multiple experts and the IT team must build up together the solution, in timely defined iterative job tasks, requiring more engagement and availability to the project.
While technology is a critical aspect, developing strong data science capabilities (deep knowledge in statistics and of the specific domain of the problem to be solved) within the healthcare organization is equally important. Today, a competent data scientist is in high demand and is not always easy to find.
In conclusion, the extensive usage of KPIs and dashboards has been promoting the adoption of BI platforms across industries and in healthcare recently. The complexities of achieving improved patient outcomes, regulatory compliance, and strategic business goals require healthcare organizations to become more proactive in the care of their patients. Current technologies have been allowing healthcare providers to execute data analytics across disparate systems running databases, data warehouses and structured or unstructured data sets, without impacting day-to-day operations or access to data. Some organizations in a more advanced stage are already applying machine learning in sophisticated diagnosis or predictive analytics.
Ultimately, healthcare providers can use information technology (Big Data or not) to optimize their analytics’ efficiency as they transition to value-based care. Moving forward, with data science based on predictive analytics, providers can deliver better patient outcomes, encounter fewer complications, decrease unnecessary emergency room interventions and experience higher levels of wellness across the population — and all at a lower cost.