In the research literature we find thousands of papers documenting academic projects documenting analytical models for the hospital environment. But the big question we need to investigate is – how much of it is actually used operationally to make a real change to treatments and patient outcomes?
As I have outlined in a previous post on hospital analytics case studies, examples of such work include queueing models used for emergency departments, patient flows, workforce requirements and what seems to be the most popular nowadays – predictive models for unplanned readmissions. Many research initiatives are also covering resource utilisation and optimisation – such as theatre utilisation, bed utilisation and so forth. However as I outlined in another post, there are many challenges in implementing and operationalising healthcare analytics, especially in hospitals.
State of a Nation
In March 2015, Jvion and HIMSS Analytics performed a survey to understand the uptake of advanced predictive modelling solutions in hospitals. For this survey, advanced predictive modelling was defined as the application of machine learning algorithms to find patterns within data to predict patient-level risks. The typical use of such analytic capabilities was targeted to assist hospital staff to prevent patient illness, avoid penalties and reduce the cost of care.
A large majority of respondents (85%) do not use advanced predictive modelling within their facilities. Of the 15% who do, 15% are academic medical centers, while 85% identify as hospital, multi-hospital, integrated delivery organisations – a large majority (88%) of these organisations have more than 100 beds in their facilities.
Respondents using advanced predictive modelling are largely applying their solutions to predict patient risk or illness (92%), with the rest using it to support other organisational goals. 82% of these predictive solutions are provided through a vendor, while the remainder have been developed in house.
Clinical goals addressed through advanced analytics include:
|
Clinical Goal |
Percentage |
| Readmissions |
27% |
| Patient deterioration |
18% |
| Sepsis |
27% |
| General patient health and need |
10% |
| Currently defining/in process |
18% |
Operational goals include the following (not rated):
- Targeting length of stay expectations
- Project reimbursements
- Target intervention activities
- Improve patient safety outcomes
- Meet nurse staffing goals
- Reduce mortality
- Reduce readmissions
- Currently defining/in process
The summary outcome of the survey is the following:
- Advanced predictive modelling solutions are gaining a foothold in the healthcare provider landscape.
- Driven by the need to improve patient health, reduce waste, and prepare for at-risk models of care, hospitals are looking to solutions that enable better health and quality outcomes while taking advantage of value-based models of reimbursement.
- This advanced analytics landscape requires a different analytic approach that can identify at-risk patients before they get sick so that interventions can be applied and penalties avoided.
In another survey of 271 healthcare professionals KPMG found that only 10% believe their organizations are using data and analytics at their highest potential and 21% are still in their “infancy” when it comes to data/analytics capabilities.
KPMG concluded that “many organizations are not where they need to be in leveraging this technology.” “Healthcare organizations need to employ new approaches to examining healthcare data to uncover patterns about cost and quality, which includes safety, to make better informed decisions.”
Although the healthcare organizations acknowledged the benefits of analytics, those surveyed by KPMG identified several hurdles to implementing the technology:
- Unstandardised data in silos (37%),
- Lack of technology infrastructure (17%), and
- Data and analytics skills gaps (15%).
What is possible?
In a recent IBM software solution brief, titled “Harnessing big data for healthcare”, which granted is aimed at selling their software solutions, IBM outlines some case studies where their solutions are used. They claim their platform provides the ability to combine clinical and research data along with environmental factors or genomic data, which then greatly assists in disease management, evidence-based medicine and better analysis and prediction of patient care and outcomes.
Some of the case studies they list include:
- Detect patients most at risk for unplanned hospital readmission, by analysing clinical and administration data along with data unlocked from doctors’ notes and patient feedback forms. Targeted action can then be taken to proactively increase the chances that patients will have a smooth transition once they leave the hospital.
- Identify patients who have a chronic illness such as diabetes or asthma.
- Analyse the history of disease progression and treatment paths among a cohort of patients with a specific diagnosis, and then work with wellness providers to proactively offer preventative programs or treatment paths for early intervention for these patients.
- Track results and the effectiveness of programs through administration and pharmaceutical data, as well as monitored patient activity to determine whether they improved or maintained their level of health, including whether they adhere to their medication and treatment program.
- Identifying patients at risk, together with their propensity to enrol in wellness programs.
Further applications of analytics can include incorporating socioeconomic factors and identifying the trends and patterns that lead to respective illnesses. And this does not even cover the use of Dr Watson to search through research material, or to assist in guided diagnoses.
Analytical maturity
Every healthcare organization is at a different point in the journey toward health analytics maturity. On a scale, health analytics maturity levels fall on a continuum starting from an ad hoc approach and moving to foundational, competitive, differentiating and breakaway levels as the organisation’s capabilities increase. As health organisations advance along this analytics maturity scale, they need increasingly sophisticated methods and technology to generate actionable insights.
IBM in their report states that “Most organizations aren’t there yet”. as illustrated in the following diagram:
Source: IBM’s report on “Harnessing big data for healthcare”.
In this report, IBM also states that “aside from government requirements, data analysis is essential to remain sustainable and cost-efficient, support clinical collaboration tools and increase access to healthcare with improved consumer engagement. Additionally, the era of big data is here, requiring organisations to manage growing volumes of structured, unstructured and streaming content.”
To create the insights that will transform the healthcare organisation, it needs a forward looking information strategy, together a platform that has proven information management and analytics capabilities, can easily integrate with the existing environment to add value, and which can scale to support new capabilities, new systems and new sources and volumes of data, including large volumes of external data. In a previous post I also illustrated how such an analytics platform can assist researchers and clinicians in a hospital setting.
Concluding remarks
The use of EHRs and conventional BI will show how you are tracking with respect to finances, patient safety and regulatory compliance, but beyond that, you’re going to need more advanced analytics to affect and improve patient outcomes.
There’s no doubt that hospitals need to improve their data analytics strategies. This is necessary, not only to perform better as value-based payment models are phased in, but more so to reduce unnecessary hospitalisations and to improve patient outcomes. Analytics are designed to reveal and understand historical patterns of data in order to predict and provide actionable insights to improve the performance of organisations. In the hospital world, operationalised use of analytics can be used to reveal at-risk patients, devise more optimal treatments, optimise the use of scarce resources such as theatres and analyse physician performance, among other insights.
