Artificial Intelligence & Medicine

Some recent articles ……

Goldberg JE, Rosenkrantz AB. Artificial Intelligence and Radiology: A Social Media Perspective. Current Problems in Diagnostic Radiology. 2018 Jul 23.
An interesting study examining the types of conversations on Twitter about the role of artificial intelligence in radiology. It appears that most tweets (or linked websites) were upbeat and wanted radiologists to continue supervision of AI-run diagnostics.

Meskó B, Hetényi G, Győrffy Z. Will artificial intelligence solve the human resource crisis in healthcare?. BMC health services research. 2018 Dec;18(1):545.
This article expresses the hope that artificial intelligence may be able to help in mitigating human resource crisis in health sector.

dos Santos DP, Giese D, Brodehl S, Chon SH, Staab W, Kleinert R, Maintz D, Baeßler B. Medical students' attitude towards artificial intelligence: a multicentre survey. European radiology. 2018 Jul 6:1-7.
This article examines the attitudes of undergraduate medical students towards artificial intelligence in radiology and medicine. Students think that artificial intelligence will revolutionize the practice of medicine, in particular radiological diagnosis.

Struggling Healthcare Organizations

A systematic review identified five characteristics of healthcare organizations that are struggling to improve quality.

1.  Poor organizational culture: Such as limited ownership, not collaborative, hierarchical, with disconnected leadership

2. Inadequate infrastructure: Limited quality improvement, staffing, information technology or resources

3. Lack of a cohesive mission: Mission conflicts with other missions, is externally motivated, poorly defined or promotes mediocrity

4. System shocks: Events such as leadership turnover, new electronic health record system or organizational scandals that detract from daily operations

5. Dysfunctional external relations with other hospitals, stakeholders, or governing bodies.

Physician Competition Effects

An interesting study examined the impact of competition among cardiologists on service provision and patients' health outcomes in the United States. Authors found that cardiologists in more concentrated markets perform more diagnostic procedures. On the other hand, higher concentration result in fewer readmissions but has no effect on mortality.

Medical Bill Collections Decrease with Patient’s Age

An interesting study that finds that the average size of medical debt decreased nearly 40 percent from when patient was twenty-seven old to when sixty-four years old. Authors thought that this was with an increase in health insurance coverage and incomes.

Well-being–some readings

Well-being has been a focus of philosophers for centuries. While philosophers tend to question how we should live, the very answer to this question begs the question how living that way will make us better, that is, how living in a certain way will make us live well.

A little self-reflection will help you to realize that there are many things that make you live well, make you happy, excited, or content, or conversely make you feel sad. Think of the things that make you feel happy. Often these things include relationships, friends, money, accomplishments. Now think of things that make you feel sad; these may include anxiety, worry, illness, poverty. Lists of both things, things that make us happy and things that make us sad, can be long and likely will vary from person to person (at least to some extent). The question one may ask is what is it that one thing (or a small group of things) that is fundamentally deterministic of feeling well.

Hedonism is a theory of well-being which focuses on individual’s pleasure or pain. Thus, pleasure is associated with high well-being and pain is associated with poor well-being. On the other hand, perfectionism focuses on our ability to develop certain virtues or characteristics. Perfectionism is similar to eudaimonia; well-being is associated with developing virtues that are human nature. Desire theory proposes that well-being is present when one gets what one desires; in other words, fulfillment of desires is associated with well-being. Objective list theories are a set of theories that have in common a list of things that make one happy; list from one theory may not overlap with another theory. This group of theory highlight the fact that it is not easy to define what constitute well-being.

Machine Learning Predicts Clinical Deterioration

Edelson and her colleagues trained a gradient boosted machine model on a dataset of almost 300K patients and were able to predict clinical deterioration better than the modified early warning (MEWS) system. Here is the link for the abstract that was presented at ATS 2018.

The simplicity of the tool which uses only age, heart rate, and respiratory is really impressive. Interestingly, respiratory rate is often the least accurately measured vital sign in the hospitals despite its simplicity. A more accurate measurement of respiratory rate can potentially help better predictive models to predict clinical worsening in a patient’s condition. Better prediction can not only help to patients by focusing on the cause of clinical worsening but will also help to more optimally utilize resources within the hospital.

Some Interesting Recent Articles from bioRxiv

1. The Subtype Specificity of Genetic Loci Associated with Stroke in 16,664 cases and 32,792 controls: Using Bayesian multinomial regression in 16,664 stroke cases and 32,792 controls of European ancestry, investigators wanted to determine the most likely combination of stroke subtypes affected for loci with published genome-wide stroke associations. Of the 16 loci, seven influenced both ischemic and hemorrhagic stroke, an EDNRA locus demonstrated opposing effects on ischemic and hemorrhagic stroke. No loci were predicted to influence all stroke subtypes in the same direction.

2. Sensitivity to Affective Touch Depends on Adult Attachment Style: Higher scores on an attachment anxiety dimension (but not an attachment avoidance) were associated with reduced pleasantness discrimination between affective vs. non-affective, neutral touch.

3. The GRE Over the Entire Range of Scores Lacks Predictive Ability for PhD Outcomes in the Biomedical Sciences: GRE scores are often assumed to be predictive of student success in graduate school; however, authors found no association in admission data.

MIMIC Database

I ran across this publicly available dataset which looks great for someone interested in data-analysis. The dataset appears to be extensive. I have not worked with its analysis yet but this will be something that will be on my ‘things-to-do” list. According to its website “MIMIC is an openly available dataset developed by the MIT Lab for Computational Physiology, comprising deidentified health data associated with ~40,000 critical care patients. It includes demographics, vital signs, laboratory tests, medications, and more.”

Proposed Changes to Medicare Physician Fee Schedule 2018

Last week, the Centers for Medicare & Medicaid Services (CMS) proposed a rule to update payment policies, payment rates, and quality provisions for services furnished under the Medicare Physician Fee Schedule (PFS) on or after January 1, 2019. Some of the proposed changes include:

1. Streamlining E&M payment with reduction in clinical documentation burden

2 .Recognition of services based on the use of communication technology

3. Creation of a bundled episode of care for substance abuse disorders

4. Flexibility on the use of radiology assistants

5. Removing functional status reporting requirement for outpatient treatment

6. A slight increase of $0.06 per RVU

7. Updating the direct practice expenses based on recent survey

8. Medicare telehealth services

9. Request for information on price transparency

10. Changes to the Quality Payment Program

Here is the link for those interested in learning more about these proposed changes

Physician White Coats and Patient Preference

An interesting study published in the BMJ Open and an interesting overview here by Brad Flansbaum.

Bottom line, formal physician attire with a white coat was rated significantly higher by patients than any other attire. Just an re-emphasis on the fact that ‘packing’ matters. One may want to to look at this that those physicians who take their profession seriously also try to wear a better representative attire than others; I am sure there a large number of physicians who may strongly disagree with this assessment.

Vitamin D Levels in African Americans

Vitamin D is associated with bone and calcium metabolism. However, more recently, studies have increasingly reported the effect of vitamin D on other organs and systems. For example, vitamin D is associated with all-cause mortality and cardiovascular mortality. Vitamin D is also associated with peripheral arterial diseases. Further, vitamin D has been associated with serum C-reactive protein levels (a biomarker of systemic inflammation). Vitamin D has also been associated with liver function.

African Americans tend to have significantly lower levels of serum vitamin D and hence may be at a higher risk of low vitamin D related morbidity. The reason perhaps lies in genetics, at least that is what this study suggests. Investigators assessed the association between percentage of European ancestry and serum vitamin D level among 2,183 African-American women. An algorithm implemented through a software was used to estimate the percentage of European ancestry versus African ancestry in each individual. After adjustment for various potential confounders (genotype batch, age, body mass index, supplemental vitamin D use, ultraviolet B radiation flux in the participant’s state of residence, and season of blood draw), investigators reported that each 10% increase in European ancestry was associated with a 0.67-ng/mL increase in serum vitamin D concentration (95% confidence interval: 0.17, 1.17).

Question remains if the low serum levels of vitamin D are clinically meaningful in African Americans. Do low serum vitamin D levels are reflective of low tissue levels (or effective tissue levels)? These type of questions will need a different and more involved study designs for answers.

Diabetes and Risk of Cancer

Patients with diabetes mellitus are at increased risk of cardiovascular complications such as heart attack, stroke, peripheral arterial disease etc. Commonly, the cause of death in diabetes is due to cardiovascular events.

Does diabetes increases risk of cancer in addition of increased risk of cardiovascular diseases? This important question was explored in this interesting study of over half-million Chinese. Investigators enrolled patients with and without diabetes and without any prior diagnosis in the study between 2004-2008 and then followed them until the end of 2013 for development of any cancer. During the follow-up period, they found more than 17,000 cases of cancer. Diabetes was associated with an increased risk of all-cause cancer and site-specific cancer.

Diabetes increased the risk of overall cancer by 13%, of liver cancer by 51%, of the pancreatic cancer by 76% and of the breast cancer by 21%. Investigators further reported that higher random blood glucose levels were associated with increased risk of cancer. Perhaps, it is the elevated blood glucose levels in diabetics that is responsible for cancer.

Medical Procedures and Heart Valve Infection

There are not free lunches! It is not uncommon to think that getting a procedure is better than not getting one, however, all procedures are associated with risks. For each procedure, the risk and benefits need to be carefully assessed before making a decision to undergo a procedure or not. Often having a procedure has greater benefits than risks, or so we assume.

While the benefits of a procedure are often clear, for example detection of coronary diseases when deciding about cardiac catheterization, the risks are not as well-studied. The list of risks associated with a procedure almost always includes risk of death and injury to some organ(s) but the actual incidence of these adverse events is usually poorly known, especially for for adverse events that are not immediately followed by the procedure (that is, there is lag-time between a procedure and associated adverse events). The more time lapses between the procedure and an adverse event, the difficult it is to ascribe adverse event to the procedure; generally so many other things, including additional procedures, happen in-between for diagnostic or therapeutic reasons.

One aspect of this issue is nicely addressed in an article published in the Journal of the American College of Cardiology where authors identified all adult patients treated for endocarditis in hospitals in Sweden and compared the occurrence of invasive medical procedures 12 weeks before endocarditis with a corresponding 12-week time period exactly 1 year earlier.

Of the 7,013 cases of infective endocarditis during the study period, several were strongly associated with cardiovascular procedures, especially coronary artery bypass grafting; procedures of the skin and management of wounds; transfusion; dialysis; bone marrow puncture; and some endoscopies, particularly bronchoscopy. A particularly interesting finding, that makes sense, is that the risk of infective endocarditis was higher if patient had a procedure while hospitalized than when patient had that procedure in outpatient setting. One way to look at this is to think that procedures should be done in outpatient setting but the converse is also possible (and more likely) that patients admitted to the hospital are sicker and hence are at higher risk of adverse effects.

Opioid use and voting Republican

A very provocative study by Goodwin et al have found that the chronic use of prescription opioid use was correlated with support for Republication candidate in 2016 Presidential elections. In case someone is concerned if the observed effect is due to the confounding by age (elderly people voting Republican but also having high need for opiates due to chronic diseases), authors did adjust for age and a broad category of other factors. I will invite everyone to read this study oneself.

Odds Ratio Limitations

Odds ratio is perhaps the most commonly reported effect size for binary outcome variables. The widespread use is likely due to the fact that logistics regression models provide odds ratio. Logistic regression models are relatively easy to implement and hence odds ratio is commonly reported. An article in JAMA recently highlighted some of the limitations of using odds ratio. The first two are well-known; the third one is not as well-known and its consequences are not fully realized. 

1) The interpretation is in odds and not in probabilities

2) Odds ratio approximates relative risk in limited situations only

3) Odds ratio depends on the the amount of unexplained variance; odds ratio may increase if the amount of unexplained variation decreases due to inclusion of strong explanatory variables in the model. Odds ratios obtained from same data set but different model may not be comparable due to the above limitation.

Productivity Paradox and EHR

Bob Wachter and Michael Howell are optimistic about the future improvements in health information technology and its impact on physicians. I share their enthusiasm and I am optimistic about an increase in productivity in healthcare sector. And why not? Its about time.

Productivity in every other sector has resulted in decreasing utilization of human resources and increasing production of goods. However, such is not the case in healthcare. A physicians still sees about the same number of patients in a day as he/she used to see 50 or 100 years ago. It is quite possible that the number patients seen might have dropped due to documentation and other requirements. The result is evident; healthcare sector now accounts for almost one-fifth of the GDP in the US. Much of this has to so with the lack of growth in productivity in healthcare sector as compared to other sectors.

What is promising is that artificial intelligence will likely to increase productivity in the healthcare. I don’t think that this will come through improvements in health information technology (HIT). I believe it will come with better utilization of HIT as data repository for learning algorithms. Technology will also help in data capture, and it does so in many ways even now, but most importantly in data inference. However, it will not be good for an average physician.

Remember, productivity can increase only if more patients can be seen by fewer physicians (or healthcare workers). That is not possible with one physician simultaneously seeing several patients. It is possible only when machines see most patients and physician sees patients only in those situations where algorithms have not gotten enough data to develop a routine.

And what those situations will occur in relatively rare conditions. These will not our usual common diseases, what we might call ‘bread-and-butter’ medicine. Rather these will be either rare diseases or rare manifestations of common diseases. A physician overseeing such a highly productive healthcare delivery system will have to be a master clinician and a superb diagnostician.

In other words, an increase in productivity in healthcare sector will spell the doom for majority of physicians ….. happy to hear what you think……….

Hospital Readmissions and Malnutrition

While it is easy to imagine how malnutrition can be related to excess readmissions, there are not very many studies examining this association. Of note, the CMS adjustment algorithm for expected readmissions takes into account malnutrition.

This retrospective study from Australia, examined this association among 26 253 admissions, representing 19 924 patients with a 6-month readmission rate of 44.8%. They found that malnourished patients had 40% higher risk of readmission within first week after discharge and 23% higher risk of readmission between the second week and 23rd week after discharge. Authors concluded that malnutrition was a strong predictor of unplanned readmissions and suggested examination of targeted nutritional interventions to lower readmission rate. 

Artificial Intelligence versus Human Doctors

While it is relatively easy to imagine artificial intelligence taking over pattern recognition type of physician work (such as reading images like radiologists or identify skin conditions by looking at the skin like dermatologists), the job of a general internist is much more demanding and requires assimilation of data inputs from a large number of sources, hence making it difficult for an artificial intelligence system to accurately predict a disease (or prepare a list of differential diagnoses).

On the other hand, the process is not, and should not be, very difficult for a well-crafted system. Razzaki et al tested such an artificial intelligence algorithm and compared it with physicians in its ability to predict diagnosis. For their study, they adopted a semi-naturalistic, role-play paradigm that simulated a realistic consultation between a patient and either their artificial intelligence system or a human doctor. Their study was designed to assess both the clinical (diagnostic and triage) accuracy and the ability to gather all of the relevant history from the patient. What they found was that their artificial intelligence algorithm was able to provide diagnostic and triage advice with a level of accuracy and safety approaching that of human doctors. If there study results are replicated then such systems may not only reduce costs and improve access to healthcare worldwide but may also provide better standardization of care delivered to patients.