How can you improve patient outcomes with data analytics?

A critical factor is the agenda of the individual or department commissioning the analysis. Too often, parameters included or excluded to skew analysis towards a pre-determined decision or outcome. Although likely benign, predetermination of outcomes can blind analysis to unforeseen consequences or unanticipated obstacles.

I learned about Dr. Hibbard's Patient Activation Measure (PAM) about ten years ago but I haven't seen it widely adopted and I don't know why. The PAM, now owned by Insignia Health, assign patients a level of engagement PAM Level 1 is "Disengaged and Overwhlemed", Level 2 is "Becoming Aware But Still Struggling", Level 3 "Taking Control and Gaining Control", and Level 4 is"Maintaining Behaviors and Pushing Further." This patient-centered approach allows clinicians to customize plans for individual patients but also at a group level. Imagine if all the newly diagnosed ESRD patients were Level 1, at an epidemiological level new ways of communicate the diagnosis might be the solution. ( I have no relationship with Insignia).

This article is very helpful to me as I am planning to do a research project on patient attendance with different illnesses to our urban clinics say over the last 3 years. Very handy tips/ tools to use.

Embracing data analytics is pivotal in shaping healthcare excellence. Beyond standard metrics, consider delving into patient-reported outcomes, which capture invaluable subjective insights. Moreover, expanding the scope to include predictive analytics could revolutionize proactive care.

The key issue is what outcomes should be measured? Do the usual outcome data improve patient safety or just operational demands? It is important to have outcomes to compare with other organisations but it is equally important that the outcomes of staff engagement, trust and psychological safety are also measured.

For those Machiavellian individuals, data collection, summary and distribution are all open to manipulation (e.g. delayed vs accelerated, or broad vs narrow distribution) such that of the spectrum of potential outcomes, those more favorable to you can be made more prominent.

An imperative component of data integration is broader data governance process within organizations. Effective data governance: 1. Creates visibility to all data assets and capabilities within the organization 2. Enables cross functional clarity, thereby reducing waste and adding value to the user 3. Protects the organization from risk by ensuring data compliance and privacy practice are maintained

What about process compliance (non-conformance) and error data? These issues are important for continuous improvement and are common in healthcare.

We are now in the era of machine learning and NLP and these can be used to analyse and visualise the datasets. We need more live data that can be visualised immediately and used to feedback into services rapidly.

Especially with the advent of AI-based decision support, a degree of competence on the parts of those responsible to validate the recommended courses of action or decisions. The power of AI is to "learn" through experience and extrapolate that learning to future analyses that may not have been in the original programming. But at least in the beginning, there will be some degree of "drift" in the recommended outcomes that have to be culled to improve the accuracy and relevance of future analyses. An analogous example can be seen in the recommendations of songs made to users of the Pandora application. Music similar to the user's defined categories are suggested, but must be validated by the user before incorporation into the playlist.

I read the other day about a retail Pharmacist chastising a pt because of their expectation that their Rx should have been ready after waiting six hours since it was called in. Should AI be used to determine staffing levels? Should the staff regularly be >more hrs behind than the pharmacy remains open? Should AI be used to send the pt a text when a new Rx or refill enters the system as to the expected time to be ready and/or texts when the anticipated time has changed for some reason?

In my opinion AI is a great tool, but that’s ultimately all it is. Especially in healthcare, I’d want human eyes on the patient and determining their care. Is there a place for AI in that mix, absolutely. But there’s no replacing the humanity that is such a huge part of medicine.

To ensure we are evaluating the right changes it is important to involve not only stakeholders but also the staff implementing the change. Change for changes sake is useless. Change which improves worker wellbeing becomes culture.

As stewards of patient well-being, data security and privacy remain non-negotiable. Embracing these challenges fuels a holistic approach, steering us toward a future where innovation thrives harmoniously with ethics, culminating in a truly patient-centered paradigm.

While data security is critical, corporate security systems are at a disadvantage due to the scale of their operations and the inertia to upgrading hardware and software (e.g. costs and training). It's important to view data security as a series of layers of protection rather than an impenetrable firewall, bearing in mind too, that a determined hacker likely has access to more advanced processing power and possesses a singular focus. Contrast that to the corporate environment with various versions of legacy hardware, multiple access points, and variable user competence. Common sense protocols should be employed, but not to the level of paranoia and paralysis of data flow.

The quality of data in healthcare is noisy and dirty. It is very difficult to use it to compare with different organisations and countries. We need international rules about which and how data should be collected into international registries to allow proper comparisons. Then we need to teach staff on the importance of keeping good data.

There are various online courses that can educate you on data analysis. As a nurse with a master's in healthcare informatics, I think it's important to have someone on your team who is a health informatics person. Having this person on your team will really help your team to ensure they are gathering and utilizing the data correctly.

Your friendly reminder that to improve patient outcomes we have to collect REaL (race, ethnicity, language) and SOGI (sexual orientation and gender identity) demographics and then talk to the communities (quantitative data is just as important). This is imperative if we are to advance health equity initiatives and improve patient outcomes for historically marginalized communities.

To piggyback off of Dr. Butler, data can be a wonderful tool in balancing out health inequities in underserved patient populations. Using data tools to gather information will help providers to discover possible solutions to the vast inequities in many underserved communities.