Health Care Realized

Measurable Progress in Health Care

Using Big Data to Improve Clinical Communication

The DNA of Clinical Communication




Last week, I had the opportunity to see two new technologies that incorporated “Big Data” and technology to improve bedside care in the hospital.




The first was part of a presentation by Michael Rothman about PeraHealth and the Rothman Index.  The work that he had done reminded me of the power of subtle indications that lie just below our threshold for perception and how, when properly harnessed, these details can help us make better decisions.

Rothman had analyzed a number of variables that appeared consistently in nursing assessments and identified which critical few had the greatest correlation with patient deterioration.   As he found correspondence among these variables, and with the overall health of the patients, he continued to add other measures such as lab values such as creatinine, and discrete parameters such as heart rate.  He pulled together all of this data and through a process called data fusion , mapped the range of each measure in a scalar way.  He then took each value and combined them to produce a number from zero to one hundred where higher values represented better health and outcomes.

The value of this synthesized number was a score that it could be trended over time to provide a graph of patient wellness.  This provided clinicians the ability to assess the condition of the patient instantaneously vs their previous level of health, and identify the velocity of change in their condition by looking at the slope of the curve.

Dr. Rothman has implemented this in a number of hospitals and seems to see consistent improvements in each.  The areas that see the greatest impact were in bolstering nursing communication, and in anticipating patient mortality to increase the utilization of palliative consults.

The other benefit was that this information could be used to improve the ability of attending physicians, or hospitalists to increase efficiency in rounding.  The dashboard that displays each patient’s RI score ranks them by severity and could assist in prioritizing the order in which they were seen.

What was particularly impressive was that his data shows the index to have 96% specificity and 93% sensitivity.  This means that the algorithm fails to accurately identify a patient who is deteriorating only 4% of the time and elicits a false alarm 7% of the time.

One challenge with the index is that it relies on nursing assessments which contain a great deal of data that is non-discrete and subjective.  It seems likely that by including 26 individual measurements, variability within some of the measurements is averaged out, but this could be a potential source of error.  Even more of a concern is the reliability of the data in some assessment measures.  As nurses are inundated with tasks, and “chart by exception”, there is tendency to cut and paste or reproduce data from previous encounters when there is no glaring deficit.  This could mask changes in patient status and decrease reliability of the score.

A confounding result may be an improvement in care associated with increased vigilance by nurses who know they are being measured.  Though the index can be a valuable tool to help nurses to validate their clinical intuition and share a quantifiable measurement with providers to direct care to some of these patients that might have previously fallen through the cracks, it does nothing to alleviate the burden on nurses of managing the overwhelming amount of data that they must manually collect and enter for a patient.



I had another meeting with folks at a large academic hospital and had an opportunity to learn about research that they were doing with a tool to drive better communication between providers, nurses, and patients and their families.

They had developed a communication platform for critical care that aggregated data and had a user interface that connected caregivers and patients.  The focus was on patient safety and measured patient condition and clinician compliance with protocols in an effort to improve in 7 areas that had been previously been identified as good targets.  The included measures were ventilator associated events, adherence to Mobility/ Ambulation protocols, maintenance of patient respect & dignity, DVT prophylaxis, alignment of goals of care, and ensuring that the team had adequately measured and made accommodations for the patient’s level of consciousness.

One of the unexpected consequences of their pilot was that patients and their families who used the touch screen in their room to communicate their understanding of the goals of care readily adopted the technology, and wanted to continue to using the device after discharge to other units.  Moreover, the researchers felt that these family members were more compliant with discharge instructions and involved in patient support when the patient arrived at home after the hospital stay.

One of the challenges that was identified was that there were discrepancies between what was manually recorded in the tool and checklists for DVT prophylaxis that were used as part of another initiative to ensure that every eligible patient received SCD therapy and heparin if not contraindicated.  The implication was that the clinicians who were measured on whether they had completed the checklist, could complete it inaccurately, to ensure that they met the requirement if time was limited.


It appeared that this solution was very effective in improving communication between nurses and providers, giving them real-time updates on how the patient was performing against these metrics and increasing the participation of the patient and family in their care.




The contrast that I found between these technologies was that while Rothman works to translate the observations of nurses into a standardized and intuitive measure, the second technology chooses a smaller number of those measurements, but distributes them in an unmodified fashion and prioritizes certain actions over others.

I found that the first technology is dependent upon a continuing and sustained effort by the nursing team, if not an improvement in their diligence around these tasks. Though I’m certain that the Rothman team does a great job with engaging nursing leadership and driving the adoption of the technology, I can imagine scenarios where implementation would place an increased burden on nurses who are already at capacity for what they can effectively accomplish.

What was illuminating about the second technology is that it actually highlighted this point.  By showing the checklists that were used successfully and led to better outcomes could still introduce errors in the record, this technology provided an insight into the fact that there can be an equal and opposite force opposing the adoption of new technology.  The second technology, to my way of thinking created a way to distribute some of the inertia of care back to the patient and their families therefore potentially reducing the workload on the clinical team.

Ultimately, both technologies show a lot of promise and have the potential to improve outcomes and drive lower costs of care.  Before considering either or similar technologies, a thorough needs assessment should be performed to determine the impact of addressing these specific aspects of clinical communication, and to ensure a thorough understanding of the saturation level of nurses’ workload.

1 Comment

  1. Great info thanks for sharing. You’re right- there are times when nurses needs quantifiable measurement to feel confident in their clinical intuition.

Leave a Reply to Cliff VanNote Cancel reply

Your email address will not be published.