Within the healthcare industry, big data is still a very generic term. Different stakeholders use it for a wide-ranging combination of clinical, genetic and genomic, outcomes, claims, social and other data that is collected from multiple sources.
These days, a lot of buzz is being witnessed around how big data can improve patient care. My take on its potential is based on a very simple premise that I often recommend. Essentially, applicability of big data in addressing potential unmet information needs in any industry is primarily dependent on a few major factors, namely,
- Accessibility of data wells
- Quality of data in data wells
- Standardization potential of the data
- Degree of separation of the data well from the beneficiary stakeholder and
- Marginal cost of data analytics
- Time utility of data analytics
- Benefit of Big Data analysis w.r.t. past prevalent data analysis
Before we move forward, I think it is important to define what a “data well” is. I have defined a “data well” to any node in the stakeholder value chain of any industry where the data is being documented and coded for some purpose. Data well is primarily applicable to already existing data pool. It is a well because often the usability (lack of structure and standardization) and ease of analysis of that data is largely not decipherable superficially. For e.g., a government run healthcare program might be capturing population healthcare indicators data on some select key indicators and coding it in paper documents stored in local administrative libraries. This is a data well. This data might be of huge importance for a pharma manufacturer. On the other hand, it might also be possible that the health indicator measured by the government run program might not be too useful (legacy issues related to program design) or the data might be too erroneous (data quality needs to be checked). Answers to all these pertinent questions cannot be gathered superficially. On a related note, efforts with digitization and standardization of EHR haven’t received desired results so far. However, mobile data could change this data landscape. We will probably discuss this topic some other day
Another aspect I want to highlight here is the definition of the term “patient” in patient care. Contrary to expectation of the term patient referring to someone who is suffering from an ailment, for big data proponents, even healthy individuals count as patients. Patients are equivalent to consumers (say, in the retail sector). Armed with the predictive analytics tools of big data, a major thrust is on how to leverage data analytics to predict and prevent disease occurrences among the population.
For reasons involving factors of priority, relevance and costs involved, initial Big Data push will be directed towards improving treatment procedures of high-cost patients and predicting their readmission. Citing the same priority and economic reasons, Big Data focus on healthy individuals (patients) will be delayed. In fact, focus on healthy patients will not at all be driven or initiated by medical practitioners. This will largely be driven by medical insurers, catalyzed by disruptive innovations from the tech start-up ecosystem (including the likes of Google and Apple) and patronized by government spending on reducing occurrences of lifestyle diseases among the population. Mobile phone data owners (Google, Apple) are at the first degree of separation with healthy patients and control data formats. Running Big Data patient health analytics is easiest for them. Insurers too have registered customer data but that is a little fragmented and lacks standardization. Technology start-ups aiding consumer with insurance management services will catalyze cloud management of patient data. Insurers will continue to collaborate with employers and build robust patient health data systems.
From the insurers’ perspective, it will also be important that behavioral health indicators that play a key role in overall physiological health management of patients are also recorded and analysed for predictions. Consequently, behavioral health, mental health and family health data (for predicting diseases of genetic origin) are soon going to be the next big buzzwords for the health insurer industry. But an obvious challenge this design would face is of privacy and data sharing issues. For patients, increase in data access for insurers would primarily mean increase in their healthcare insurance costs. Whether that is a myth or not, we will discuss is a separate topic and note.
To assess how big data can help patient care, we need to look at the types of data-sets that a hospital entity can gather on its patients.
- Patient historical treatment data and current bedside monitoring
- Hospital historical patient data (diagnosis, treatment, outcomes, learning)
- Hospital historical clinical care processes data (staffing, bedside care and patient transfer)
- After discharge patient data (readmission and post discharge care) and outcomes
- External (other hospitals, physicians, clinical trials) data on patient treatment, relevant drug prescription and outcomes
- Historical hospital data on intensive care treatments (with potential of multiple organ failure)
- Geo-spatial or family data on patients to decipher likely diagnosis and treatment outcomes
As I stated earlier, first Big Data interventions will be seen in improving patient care of high cost patients. One of the key trends we will witness is utilization of real time bedside monitoring data. Real time monitoring of multiple data streams bedside data to keep a tab on patient’s health status (physiological data tracking) is going to be one of the foremost adoption areas of Big Data. Bedside monitoring data, when coupled with other processed data, could also help hospitals optimize and prioritize its clinical workflow including managing staff, patient transfer and bedside care management. Developing predictive algorithms, based on gathered data on high cost patients, to predict readmission probability can help building targeted interventions to prevent such instances and improve post discharge care plans. Hospitals will also leverage its historical patient data treatment registries to predict potential diagnosis and preventive care of its new patients. If the expanse of the data registries analysed could be widened further and shared between hospitals and other stakeholders, management of treatment for patients with chronic conditions and potential of multiple organ failure in intensive care can be significantly improved. Gathering external data on prescription drugs and changes in vital signs of patients could help preventive care further.
The biggest challenge, as we see today, is the lack of skilled professionals in Big Data.