None are the singular cause of the jaw-dropping financial inefficiencies of the health care industry. But together they create a cost curve that must be bent downward, and to do so requires starting with the low-hanging fruit where, without decreasing quality or access, caregivers can still decrease cost.
Hospital readmission is plainly one of those areas. Especially in an era of aging patients where chronic incurable conditions like diabetes or heart disease are profligate, the best case for a patient is to stabilize their condition, keep them out of the hospital and allow them to live as comfortably as they can.
“Cost, safety, collaboration, convenience are all reasons why we want to control readmissions,” says Dr. James Holly, M.D., CEO of Southeast Texas Medical Associates. Reducing readmission became a central goal at SEMTA, and through the use of electronic records and analytics, the 29-physician practice in Beaumont achieved a 22 percent improvement over six months.
Clusters of Data
Holly sees many geriatric patients who might have seven or eight ongoing conditions. The success in treating each one is gauged by a set of outcome metrics by which the level of care and the patient’s response can be compared with standards derived in medical research.
The American Medical Association’s Physician Consortium for Performance Improvement (PCPI) is one such source of a quality metric set covering various conditions that allows a provider to measure their own performance at the point of care. Other sources for quality metric sets include the National Committee for Quality Assurance (NCQA) and the Ambulatory Quality Association (AQA).
Where process metrics measure the degree to which a clinical best practice was followed, outcome metrics are harder to quantify, because outcomes are judged by an assortment of health and quality of life measures, many of which are subjective. These include “hard” metrics like readmission rates, but also freedom of movement or lack of pain, satisfaction and engagement.
“We discovered that if you’re only tracking only one quality metric about a particular condition, it’s really not going to change anything,” says Holly. “But if you’re tracking a cluster of seven or eight quality metrics about a specific condition and you hit the metric on those, you’re very likely going to be changing the outcome for patient.”
Across whatever range of conditions a typical patient might have, SEMTA physicians might assemble as many as 50 or 60 quality metrics, a galaxy of multiple clusters. If a patient’s progress can be managed across that many metrics, evidence shows it will change the trajectory and outcomes of that patient's care.
Data Sets and Deviations
Doctors using medical records to manage patient outcomes have traditionally had to look back at what happened 12 or 18 months earlier to determine their success. Getting ahead of that time curve to address the current health of a patient calls for an analytical approach that suggests actions to take today or tomorrow.
Tracking the course of multiple conditions in a single patient requires tools that mine data for different themes, modes, medians and standard deviations. Holly is a big proponent of statistical analysis and using standard deviation to calculate patients’ – and care providers’ variance from mean performance – based on statistical evidence. Southeast Texas’ standard deviation for diabetes fell from 1.98 in 2000 to 1.2 in 2010, for example. Statistically, Holly says, that took the practice “from terrible to much better, not perfect but closer to the .9 deviation you might expect in human biology.”
It’s a relative measure, but that’s the point, Holly has found. “There's a thing in health care called clinical inertia where a patient comes in, they are not at their care goals, their blood pressure or lipids are not treated as well as they could be, and yet nothing is done.” Research suggests that clinical inertia, defined as lack of treatment intensification in a patient not at evidence-based goals for care, is frequently a cause for preventable errors.
To counter widespread clinical inertia, Holly and SEMTA publicly reported its own doctors' and nurse practitioners' performance for different disorders. The data is posted by populations and not by patient name (though patients are given the information on their own quality of care at every visit).
When such information is posted, Holly says, “The first complaint you hear from providers is, ‘the data is wrong, I do a better job than that.’” The need to back reports with defensible data led to an “exhaustive” business intelligence implementation that involved Cognos software and IBM business partner LCI. For doctors to buy in, the methodology for auditing and analysis had to be bulletproof as well.
Southeast Texas analysts initially authenticate data with chart reviews and hand checking to confirm the correlation, which Holly says became “spot-on excellent” for the purpose. “Once providers understand the data is accurate and is being publicly reported, it really stimulates them to pay attention to each individual patient. We spent a lot of money and time adapting Cognos to health care to get to that, but when you take care of each individual you'll be taking care of a panel or population in the process that supports that.”
Different patient populations were compared with those who came back as scheduled or not, and whether (and how) that led to readmission. It looked at co-morbidities, secondary or associated symptoms patients had and whether that affected their return. Readmission was checked by demographics of ethnicity, gender, age, insurance or lack thereof. Analysts then drilled down to look for small process “levers” or indicators in health care standards and practices that might suggest a minor intervention that would reduce the likelihood of readmission.