Electronic health records: Useful information or bad ‘aligatorithms’?

Electronic health records (EHR) were expected to banish medical errors from bad handwriting, save money, enable interoperability between health-care providers, and create vast data of patient care in which Artificial Intelligence (AI) and algorithms could run to assist with diagnoses.

In 2008, about 9 percent of the health-care industry was cautiously trying, testing, evaluating and selecting EHR systems, a process likely gradually to allow the best systems to surface as industry standard. Seduced by rosy expectations and thinking it knows best, the federal government mandated in 2009 that health-care providers adopt EHR forthwith. To make this happen, the government offered to pay some $36 billion in incentives to providers who could demonstrate that they had promptly adopted and were “meaningfully” using a government certified electronic health record technology. Penalties loomed for tardy providers.

The rush was on. Technology companies scrambled to make their own proprietary certified EHR software with little input from health-care providers and with the attitude that security and other flaws could be fixed later. Providers scrambled to buy and implement their products so that they could meet deadlines for government incentive payments. What has been the result?

▪ First, there has been some improvement in patient safety, but off-setting mistakes arise from clunky, impractical software. Examples abound. If an emergency room doctor seeks to order Tylenol, he/she can find a drop-down screen with 86 options, many of which are irrelevant to the patient. The harried doctor can easily click on a neighboring wrong item in scanning down to Tylenol. EHRs have been found at some institutions to accurately match the record to the patient only 50 percent of the time. When ibuprofen is prescribed for a female, the prescription will be blocked by a pregnancy warning, which must be overridden by more clicks. As a health professional clicks and zooms in and out of the record to override an irrelevant warning, do an order or find information, one can bounce into the wrong record. Nonmedical personnel designing the software was obviously contraindicated.

▪ Second, health-care providers compete with each other for business. Like any other business, they seek to protect their confidential proprietary information. Ongoing resistance to interoperability between competitors was foreseeable but not accounted for. Coke cannot be forced to share its formula with Pepsi.

▪ Third, in August 2014, the Department of Health and Human Services’ Office of the Inspector General found that for four out of five EHR certifying bodies reviewed, the government had failed to ensure that testers were properly trained and knowledgeable about the security-related requirements they were testing. Hacking of vulnerable health-care institutions and ransomware attacks have snowballed.

▪ Fourth, the promise of predictability from Big Data is apparently being fulfilled. Ranging over vast patient data, AI algorithms have been found to achieve 85 percent accuracy in diagnosing gastrointestinal diseases in children and 93 percent for bacterial meningitis. Algorithms are also able with extraordinary accuracy to predict and diagnose other diseases, such as cardiovascular illnesses or cancer. In one study, within 24 hours of a hospital admission, algorithms predicted with 90 percent accuracy the patient’s odds of dying. From a caller’s voice, words and background noise, software has been found to correctly detect cardiac arrest in 93 percent of cases versus 73 percent for human dispatchers. But, researchers are unable to explain or audit how algorithms work.

▪ Fifth, one 2016 study found that about half of the physician’s time with a patient is spent facing a computer screen to do electronic tasks. About two hours of computer work are required for every patient hour. A 2018 Merritt Hawkins survey indicates that 78 percent of physicians have symptoms of burnout in what the Harvard School of Public Health calls a public health crisis.

▪ Sixth, from my own experience as a health-care attorney, a request for copies of medical records can result in reams of confusing printouts of disjointed EHR-formatted forms. Electronic records are highly vulnerable to hard-to-detect fraudulent changes, but crucial auditing functions can be inadequate, not enabled or too easily disabled.

For better or worse, EHR is here to stay. So are slam-downs of ill-conceived government mandates unless policy-makers become better educated. With Tarzan nowhere in sight, let’s hope that “aligatorithms” are not lurking under the surface of indecipherable algorithms.