Forcing Quality — is “big data” the answer?


Today, 7/12/13, the Wall Street Journal has a front page story “Hospitals Prescribe Big Data to Track Doctors at Work” by Anna Wilde Mathews.  The past 20 years have seen numerous efforts to give performance data to doctors with the hope the doctors will change practice patterns.  Each time the efforts have failed.  The Wall Street Journal article  paints a hopeful picture that faster and more accurate data analysis will finally solve health care quality problems.

Insanity: doing the same thing over and over again and expecting different results.                                    Albert Einstein

Clearly, collecting such data does reveal quality problems.  In fact, the first step to correcting quality problems is finding them.

But, there are three fatal flaws in thinking that telling doctors about their own quality problems will really help patients:

  1. Studies of errors happening with complex tasks show the best human performance is an error rate of 1 – 10 errors  per 100 tasks.  Meaning, no matter how much you flog doctors with data there is a human limit to performance.  Patients want performance in the range of 1 error per million which requires computers and systems.
  2. By looking at errors as a personal failing instead of a system failure innovation is inhibited.  People tend to say “try harder” rather than “try something new”.
  3. A data driven focus on past problems actually blurs the view of new solutions and new treatments.  Decision support can assist physicians to adopt new methods and treatments.  Currently it takes about 15 years for proven treatments to enter routine practice — big data does not move that forward.

The electronic medical record (EMR) solves many of the problems (this is not big data but something called decision support). The big data approach is to tell doctors they failed and to try harder in the future.   Instead, decision support shows doctors choices at the point of making a decision. For example, as the physician is using the EMR with a patient in the room these messages could show up:

Mammograms: the EMR tells the physician that mammograms will be ordered every year unless you check here [ ].

Tetanus vaccination: your patient has not had a tetanus vaccination.  Give the vaccination now? [ ]yes [ ] no

Asthma medications: Refill data suggests the patient is not taking enough controller medication. What is the reason? ___________________.

Chlamydia screening: Your patient (age 16-24) has not had screening. Do it today? [ ]yes [ ] no

Bronchitis treatment: Treatment guidelines for acute bronchitis do not include antibiotics. Is the diagnosis correct?.   Consider other agents like cough suppressants or bronchodilators.

Diabetes & blood sugar. Guidelines suggest checking A1C hemoglobin every 3 months for diabetics taking insulin.   Order A1C now? [ ]yes [ ]no

 Conclusion:  It is important to collect data on individual doctors and on hospital systems.   That data can tell whether quality improvement efforts are working.  But, doctors  need computer driven decision support at the time of ordering treatment or tests, not criticism days or months later.  Patients benefit immediately from decision support but only later, if at all, from big data.

, , , , ,

%d bloggers like this: