Bio-Medical Computing(6.872/HST.950)

Peter Szolovits, PhD (psz@mit.edu) Gil Alterovitz, PhD (gil@mit.edu) + guest lecturers TA: Tristan Naumann (tjn@mit.edu) Course Secretary: Fern Keniston (fernd@mit.edu)

https://stellar.mit.edu/S/course/6/fa15/6.872J

Mon, Wed 9:30-11:00am, 36-155

Who Am I?

• Background is in Artificial Intelligence (AI)• Got interested in medicine as an application of AI, in 1974• Knowledge-based systems for diagnosis, therapy selection & management• Early emphases

• Reasoning based on causal, pathophysiological models• time, space and probability

• 1990’s revolution brought tons of data; e.g., MIMIC contains terabytes of ICU data• Shift from knowledge-based to empirical models• Data mining and machine learning: logistic regression, naive Bayes, Bayes

networks, neural nets, support vector machines, conditional random fields, etc.• Challenge: combining general knowledge with experience

• Natural Language Processing to enable use of clinical data locked up in narratives• Shift from institutional to personal systems: Guardian Angel (http://ga.org)• NRC studies on patient privacy, health care challenges to IT, CMS IT revitalization

Why do People Care?

• “The Learning Health Care System” — Institute of Medicine• Growing availability of clinical data on past patient care

• E.g., Partners has a Research Patient Data Registry (RPDR) containing records of ~4M patients

• E.g., Kaiser Permanente has (2012) ~32 petabytes of data (much of it imaging)• 2009 Stimulus Bill subsidizes hospitals, clinics, doctors’ offices to install

electronic health records to meet “meaningful use” criteria• Improving machine learning methods to exploit such data

• “The $1000 Genome” — spread of genomic medicine• Phenotype = f(Genotype, Environment); what is f?• Theories of genetic causation: (a) Mendelian, (b) Common Variants, (c) Rare

Variants, (d) neo-Mendelian, (e) ???• Cost

• E.g., a Chevy contains more in employee health care costs than steel!

Inexorable Growth of Healthcare Spending

Healthcare Spending as % of GDP

0

6

12

18

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

2012

USA France UK Taiwan

The Challenge of Productivity

• Productivity = goods & services produced / resources consumed• Staying in place is not good enough

0%

10%

20%

30%

40%

50%

60%

70%

80%

1840 1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010

farm labor (% of population)

0

10

20

30

40

50

60

70

80

1840 1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010

Farm productivity

What is the Productivity of Health Care?

• Hard to know what outputs to measure• Life expectancy• Childhood mortality• Quality of life

• Difficult to estimate resources consumed

Comparison of 1901 and 2003 survival curves for US white women.From http://www.state.gov/g/oes/rls/or/81537.htm

0"10000"20000"30000"40000"50000"60000"70000"80000"90000"

100000"

0" 5" 10"

15"

20"

25"

30"

35"

40"

45"

50"

55"

60"

65"

70"

75"

80"

85"

90"

95"

100"

105"

110"

115"

MODEL based on real 2004 mortality data

UK

Standing Still is Not Good Enough

Chris Dede, Harvard Ed

http://theincidentaleconomist.com/wordpress/the-health-care-productivity-problem/

Less Efficient Sectors Come to Dominate

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Mining

Construction

Other services

Health, education

Arts

Prof. services

Finance

Transport

Nondurable

Utilities

Retail trade

Wholesale trade

Argiculture

Information

Durable goods

Hypothetical sector growth over 30 years, assuming constant demand for each sector. Productivity rates from OECD 2009.

“The Learning Health Care System”

• “one in which progress in science, informatics, and care culture align to generate new knowledge as an ongoing, natural by-product of the care experience, and seamlessly refine and deliver best practices for continuous improvement in health and health care” —IOM

• Needs not currently met:• Comprehensive collation of all clinical, social, demographic, behavioral, ... data

that are now captured in the health care system• Routine capture of novel data sources:

• genomes, gene expression, etc.• environmental factors (e.g., metagenomics)• physiological response to life situations

• (related to fitness and wellness)• Technical infrastructure

• Storage and analysis of truly “big data”• Incentives and demonstrations of utility

9

What is Evidence-Based Medicine?

• Randomized Controlled Clinical Trials• E.g., is drug A more effective than drug B for condition X?• Narrow selection of patient cases and controls• Careful collection of systematically organized data• Statistical analysis of outcomes• => Statistically significant conclusions

• But:• Heterogeneity: Most cases to which RCT results are applied do not fit trial criteria• Short Follow-Up: Trials run for limited times, but use is longer• Small Samples: Some effects are rare but devastating

• Instead: consider every patient’s experience as a source of knowledge by which to improve health care

Vioxx and Heart Attacks

time series of MI-related hospitalizations by Poisson regressionwhere the two drugs were included as indicator variables. Themodel for rofecoxib revealed a positive relationship with counts ofMIs (rate ratio 1.14; 95 percent confidence interval, 1.09 to 1.19;P,0.001), explaining 28.7 percent of the deviance. The model forcelecoxib also revealed a significant positive relationship withcounts of MIs (rate ratio 1.09; 95 percent confidence interval, 1.01to 1.18; P= 0.022), explaining 5.2 percent of the deviance.Finally, we measured the yearly impact of estimated celecoxib

and rofecoxib use on hospitalizations due to MIs as a dose-response relationship using national prescription data (Table 1).Prescriptions for celecoxib and rofecoxib peaked in 2001 with25.41 and 27.06 million prescriptions dispensed nationally,respectively. Incidence of MI-related hospitalizations also peakedin 2001 with 62.3 hospitalizations per 10,000 (Figure 2). Yearlycounts of MIs were significantly predicted by prescriptions forrofecoxib (0.7% increase in MIs per million prescriptions; 95percent confidence interval, 0.2% to 1.3%) and marginally forcelecoxib (0.6% increase in MIs per million prescriptions, 95percent confidence interval, 20.005% to 1.8%;). For every millionprescriptions of both rofecoxib and celecoxib, we found a 0.5percent increase in MI (95 percent confidence interval, 0.1 to 0.9),explaining approximately 50.3 percent of the deviance in yearlyvariation of MI-related hospitalizations.We also find a significant negative association between mean

patient age and prescriptions for celecoxib and rofecoxib(Spearman correlation, 20.67, P,0.05). Over the years whenrofecoxib was not available, the mean age was 66.5 compared toa mean age of 65.4 over the period of market availability. Figure 3

shows the nadir in age corresponding to the peak of rofecoxib andcelecoxib prescriptions.

Anchor DiagnosesTo demonstrate that our findings were not simply based onchanging healthcare utilization, we investigated temporal trendsfor hospitalizations for three other diagnoses. Though showinga positive linear increase, abdominal aortic aneurysm (Spearmancorrelation, 0.17; P = 0.66), appendicitis (Spearman correlation,0.11; P= 0.78), and pneumonia and influenza (Spearmancorrelation, 0.071; P= 0.86) were not associated with trends inMI. Further, only trends in inpatient visits for MI weresignificantly correlated with national prescription rates forrofecoxib and celecoxib (Spearman correlation 0.93; P,0.001).

DISCUSSIONOur findings demonstrate striking longitudinal trends in thepopulation rate of MI and strongly suggest a population-levelimpact of rofecoxib and celecoxib on MI. Temporal trends ininpatient stays due to MIs are tightly coupled to the rise and fall ofprescription of COX-2 inhibitors, with an 18.5 percent increase ininpatient stays for MI during the time when rofecoxib was on themarket. These results are consistent with prior reports demon-strating cardiovascular effects of rofecoxib and celecoxib.[9,10,11,12,13,14,15,17,27] Further, mean age at MI appears tohave been lowered by use of these medications.Using population surveillance methods [28,29,30] we confirm

a sharp rise in adverse cardiovascular events within eight months

Figure 1. Cumulative sum (CUSUM) chart of monthly incidence of hospitalizations due to myocardial infarction from January 1, 1997 to March30, 2006. Assigning 1997 and 1998 as a baseline period, a target mean of 47.1 hospitalizations due to myocardial infarctions per 100,000 anda standard deviation of 2.8 per 100,000.The threshold for CUSUM was calculated as 1.18, yielding an average-run-length of 50 (dashed red line). Anaberration, two standard deviations from baseline, was initially detected in January of 2000, with a CUSUM value of 1.37 and 51.5 myocardialinfarction-related hospitalization per 100,000 stays (solid line).doi:10.1371/journal.pone.0000840.g001

COX-2 Inhibitors and MI

PLoS ONE | www.plosone.org 3 September 2007 | Issue 9 | e840

“Trends in inpatient stay due to MI were tightly coupled to the rise and fall of prescriptions of COX-2 inhibitors, with an 18.5% increase in inpatient stays for MI when both rofecoxib and celecoxib were on the market (P<0.001). For every million prescriptions of rofecoxib and celecoxib, there was a 0.5% increase in MI (95%CI 0.1 to 0.9) explaining 50.3% of the deviance in yearly variation of MI-related hospitalizations.”

Brownstein JS, Sordo M, Kohane IS, Mandl KD (2007) The Tell-Tale Heart: Population-Based Surveillance Reveals an Association of Rofecoxib and Celecoxib with Myocardial Infarction. PLoS ONE 2(9): e840.

Adoption of Electronic Health Records

• Rapid adoption in hospitals, likely driven by 2009 “Stimulus Bill”• “Basic EHR systems include …:

patient demographics, patient problem lists, patient medication histories, clinical notes, electronic orders for prescriptions, laboratory results viewing, and imaging results viewing.” —ONC(Data from http://dashboard.healthit.gov/HITAdoption/?view=0)

• Positive factors for adoption: large size, urban location and HMO penetrationHealth information technology adoption in U.S. Acute care hospitals. Zhang NJ, et al. J Med Syst. 2013 Apr;37(2):9907

• … some reasons for suspicion about adoption rates

https://www.healthit.gov/sites/default/files/oncdatabrief16.pdf

Adoption of Electronic Health Records EHR Adoption in Doctors’ Practices

14

% A

dopt

ion

0

10

20

30

40

2008 2009 2010 2011

Overall Primary Care Small PracticesRural Practices

http://dashboard.healthit.gov/HITAdoption/?view=0

E-Prescribing Number of Users of Surescripts Network

15

0

150000

300000

450000

600000

2008

-12

2009

-220

09-4

2009

-620

09-8

2009

-10

2009

-12

2010

-220

10-4

2010

-620

10-8

2010

-10

2010

-12

2011

-220

11-4

2011

-620

11-8

2011

-10

2011

-12

2012

-220

12-4

2012

-620

12-8

2012

-10

2012

-12

2013

-2

Providers E-Prescribing Use EHR to E-PrescribeStandalone Systems

0

25

50

75

100

2008

-12

2009

-820

10-4

2010

-12

2011

-820

12-4

2012

-12

% Pharmacies

16

Universal Adoption of Electronic Health Records• Problems of Interoperability

• Standards vs. evolution• Minimalist: HL7-CCD or CCR: text + optional codes (SNOMED, LOINC, …)• Comprehensive data: HL7-RIM• Instance of Berners-Lee’s Semantic Web issues

• Poor health information exchange• Workflow changes

• Errors• E.g., mortality 2.8% ⇒ 6.6% after CPOE system @ Pitt Children’s

• Waste of effort• “Cut and Paste” records

• Proprietary lock-in

Han et al. Unexpected increased mortality after implementation of a commercially sold computerized physician order entry system. Pediatrics (2005) vol. 116 (6) pp. 1506-12

Meaningful Use for Quality Improvement

• 2009 Recovery Act defines “Meaningful Use”:• Use EHR in a meaningful manner (e.g., e-prescribing)• Electronic exchange of health information to improve quality• Submit clinical quality measures

18

Stage 1 Data Capture & Sharing

2011-12

Stage 2: Advance Clinical Processes

2014

Stage 3: Improved Outcomes

2016

Electronically capturing health information in a standardized format

More rigorous health information exchange (HIE)

Improving quality, safety, and efficiency, leading to improved health outcomes

Using that information to track key clinical conditions

Increased requirements for e-prescribing and incorporating lab results

Decision support for national high-priority conditions

Communicating that information for care coordination processes

Electronic transmission of patient care summaries across multiple settings

Patient access to self-management tools

Initiating the reporting of clinical quality measures and public health information

More patient-controlled data Access to comprehensive patient data through patient-centered HIE

Using information to engage patients and their families in their care

Improving population health

http://www.healthit.gov/policy-researchers-implementers/meaningful-use

Reducing Readmissions

• Starting Oct 1, 2012, hospitals are penalized by Medicare if a patient is re-admitted for the same condition within 30 days• Initial focus conditions: Acute Myocardial Infarction, Heart Failure, Pneumonia• Risk adjustments for demographics, co-morbidities, patient frailty

• Predictive models are worth $$$• E.g., Potentially avoidable 30-day hospital readmissions in medical patients:

derivation and validation of a prediction model. Donzé, et al., JAMA Intern Med. 2013 Apr 22;173(8):632-8• Retrospective study of 10,731 discharges from Partners HealthCare• 2,398 (22.3%) readmitted within 30 days, 879 (8.5%) judged avoidable• Eliminated planned re-admissions and those for other conditions• Logistic regression model predicts avoidable re-admission with AUC=0.71,

good calibration• Important variables: hemoglobin at discharge, oncology service, Na+ level at

discharge, procedures performed, non-elective admission, number of admissions in past 12 months, length of stay

19

118 papers on “predict readmission” since

2010 in Pubmed

Medical Informatics

• Intersection of medicine and computing • Plus theory and experience specific to this

combination • =Medical Computing, ~Health Informatics

• Science • Applied science • Engineering

Types of Bio-Medical Informatics

• Cellular level: Bioinformatics, Systems Biology

• Patient level: Clinical Informatics, Health I., Medical I., …

• Population level: Public Health I. • Imaging Informatics

Bio-Medical Informatics

• Phenotype = Genotype + Environment • In humans, we rely on “natural

experiments” • Measurements

– Genotype: sequencing, gene chips, proteomics, etc.

– Environment: longitudinal surveys, etc. – Phenotype: clinical records, assembled to

longitudinal data

Class Outline

• Sources of Data – Clinical

• signs, symptoms, diagnoses, labs, meds, procedures, bedside measurements, monitored signals, notes, discharge summaries, …

– Environmental • air quality, toxicology, disease outbreaks, …

– Genomic • chromosome abnormalities, SNPs, CNVs, …,

exome sequences, full genome sequences

• Methods of modeling 23

Data and Models

• Sources of Data• Clinical

• signs, symptoms, diagnoses, labs, meds, procedures, bedside measurements, monitored signals, notes, discharge summaries, …

• Environmental• air quality, toxicology, disease outbreaks, …

• Genomic• chromosome abnormalities, SNPs, CNVs, …, exome sequences, full genome

sequences• Methods of modeling

• predictive models• supervised & unsupervised methods• causality• pitfalls

Real World Applications

• Practice• Clinical decision support• Epidemiology and public health• Personalized life-long health care & prevention• Telemedicine and mobile health

• especially in the developing world• Undiagnosed diseases• Cancer• Intensive care

• Institutional• Workflow• Management• Standards and interoperation• Commercialization

The Medical Cycle

interpret

plan

formulatepatient

datainformation

diagnosistherapy

initial presentation

Care Processes

• Data: instrumentation, monitoring, telemetry • Information: interpretation, filtering, sampling,

smoothing, clustering • Diagnosis: inference, model-based reasoning,

classification • Prognosis: prediction, natural course,

experience • Therapy: planning, predicting effects,

anticipating

Meta-level processes

• Acquisition and application of knowledge • Education • Quality control and process improvement • Cost containment • Reference (library)

Settlement

Membership

Care Team

Int. Med

Rad

Lab

Pharm

Surgery

Account

Plan Design

Discharge Dismiss

Episode Activation Authorization

Health Status Mgt.

Community Care

Self Care

Evaluate

Assess

Plan

Act

Schedule

Refer Visit Activation

Mgt. Team

Health Mgt.

Health Record

Measure

Enterprise-level Clinical Process Automation...

figure from David Margulies

The “Learning Health Care System”

Membership

Care Team

Int. Med

Rad

Lab

Pharm

Surgery

Account

Plan Design

Discharge Dismiss

Episode Activation Authorization

Health Status Mgt.

Community Care

Self Care

Evaluate

Assess

Plan

Act

Schedule

Refer Visit Activation

Mgt. Team

Health Mgt.

Health Record

MeasureObserve/Measure

Analyze

Model Plan Intervene

•Process •Medical content

Dogma

Phenotype = f( Genotype , Environment)Traits Diseases Behaviors …

Gene sequence SNP’s Expression data …

Diet, smoking, drugs, … Insults and injuries Exposures …

• What is the functional form? • How do we investigate these relationships? • Can we take advantage of the exponential

growth of genomic data?

Growth in Gene Expression Omnibus Measurements

2004 ~30,000

2010 472,929

2014 987,662

2015 1,569,998

http://ncbi.nlm.nih.gov/geo/

Where are the Phenotype and Environment-related Data?

Phenotype = Genotype + Environment

Traits Diseases Behaviors …

Gene sequence SNP’s Expression data …

Diet, smoking, drugs, … Insults and injuries Exposures …

• Perform Controlled Experiments? – Unethical using human subjects!!! – OK on rats.

Experimental Subjects

High-throughput phenotyping at Medical College of Wisconsin

Where are the Phenotype and Environment-related Data?

• Environment – (Hardest to get) – Questionnaires,

• e.g., Nurses’ Health Study, Framingham Heart Study

– Monitoring • e.g., LDS hospital infectious disease monitors

– Metagenomics • DNA of the stuff living in your guts

• Phenotype – “Natural Experiments” – ∴ Clinical Data

Consent

Banking Genomics

Annotation

Family Hx

The fantasy: Informatics for Integrating Biology & Bedside

I2b2: I. Kohane, et al.

Plausibility• Phenome-Genome Network

– Gene Expression Omnibus • expression data • annotations: tissue, disease,

expt. conditions, … – Interpret annotations to UMLS – Differential expression vs.

condition – Interesting relations:

• 11 genes & aging • DDX24 and leukemia • 2 genes & injury

Butte & Kohane, Nature Biotech 2006

Clinico-Genomic Research• Identify a highly specific clinical population, and

controls • Gene-wide association studies (GWAS) • Hope that notable differences appear between those

with/those without disease • Disease models:

– Mendelian – Single-nucleotide polymorphisms – Private variation – ?

• xWAS: Phenome, Environmentalome, …39

Time scale in medicine

• Cure—usually acute illness • Manage—long-term, chronic illness • Prevent • Predict (especially based on genetics)

WHO Constitution defines “health”

“a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity”

• Physical • Mental • Social —very hard to measure

Distribution of Ages

• Life tabledeaths by year(Japan, 1989)

Life table death rates by age

0.0001

0.0010

0.0100

0.1000

1.0000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

male p(death)female p(death)

US SSA 1997

Life table cohort survival

0

25,000

50,000

75,000

100,000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

males alivefemales alive

US SSA 1997

Measures of Health

• Longevity at birth (CIA World Fact Book, 2001)Country Male Female 2015Rwanda 38.35 39.65 59.26Kenya 46.57 48.44 63.52South Africa 47.64 48.56 49.56Cambodia 54.62 59.12 63.78Brazil 58.96 67.73 73.28Russia 62.12 72.83 70.16Albania 69.01 74.87 77.96USA 74.37 80.05 79.56Japan 77.62 84.15 84.46

Causes of death(industrialized countries, 1989)

Circulatory system

48%

Malignant neoplasms

19%

Accidents 7%

Others 26%

Quality of life

• Value of a total life depends on – Length (assume now is N) – Quality (q) over time – Discounts (g) for future or past

(depends very much on what the value is to be used for)

Modeling life quality

Mortality, Disability, Morbidity

0"10000"20000"30000"40000"50000"60000"70000"80000"90000"

100000"

0" 5" 10"

15"

20"

25"

30"

35"

40"

45"

50"

55"

60"

65"

70"

75"

80"

85"

90"

95"

100"

105"

110"

115"

0"

10000"

20000"

30000"

40000"

50000"

60000"

70000"

80000"

90000"

100000"

0" 5" 10"

15"

20"

25"

30"

35"

40"

45"

50"

55"

60"

65"

70"

75"

80"

85"

90"

95"

100"

105"

110"

115"

0"

10000"

20000"

30000"

40000"

50000"

60000"

70000"

80000"

90000"

100000"

0" 5" 10"

15"

20"

25"

30"

35"

40"

45"

50"

55"

60"

65"

70"

75"

80"

85"

90"

95"

100"

105"

110"

115"

0"10000"20000"30000"40000"50000"60000"70000"80000"90000"

100000"

0" 5" 10"

15"

20"

25"

30"

35"

40"

45"

50"

55"

60"

65"

70"

75"

80"

85"

90"

95"

100"

105"

110"

115"

Top 10 Chronic ConditionsPersons aged ≥ 65

Condition Both Male FemaleArthritis 49.6 40.7 55.7Hypertension 39.0 33.0 43.2Hearing impairment 30.0 35.2 26.3Heart disease 25.7 26.9 24.9Orthostatic impairment 16.8 15.7 17.8Cataracts 15.5 11.3 18.4Chronic sinusitis 15.2 13.7 16.2Visual impairment 10.1 12.0 8.8Genitourinary 9.9 11.3 8.9Diabetes 8.9 7.8 9.7

U.S. Nat’l Ctr Health Stat, Vital and Health Statistics, 1985 (1982 data)

Societal quality of life

• Aggregation of individual qualities + Equity (distributions)

• Is more better? (Population control.) • Is less better? • How much to spend?

Who makes decisions?“In those days there was no bureaucratic regimentation, there were few forms to fill out, malpractice premiums were affordable, and the overhead costs of running a practice were reasonable. Our bills were simple, spelled out so anybody could understand them without the use of codes. Patients usually paid their own bills, promptly too, for which an ordinary receipt was given. Hospital charges were set by the day, not by the aspirin. Medical care was affordable to the average person with rates set by the laws of the marketplace, and care was made available to all who requested it regardless of ability to pay. Doctors were well respected; rarely were we denigrated by a hostile press for political reasons. Yes, in the days before government intervention into the practice of medicine, doctor’s fees were low, but the rewards were rich; those were truly the ‘golden years’ for medicine.”

Edward Annis, past President of AMA Code Blue, 1993

Aggregation

• Trend: social aggregation leads to decisions at a larger scale – Multi-specialty providers – Government guarantees and mandates – Risk sharing – Oregon-wide spending “optimization”; – British NHS

Changing Context of Health Care

• Fee-for-service • CMS (Center for Medicare and Medicaid

Services pays for Medicare, ~40% of US $ • Capitation

– HMO’s (Health Maintenance Organizations) take overall responsibility to care for patient for fixed fee

– Pushing risk down to the physician or group

Determining Factors:

$£€¥RExponentially growing expense

of health care• More healthcare than steel in GM cars • Increased demand

– Much more possible – Better tests, therapies – High human motivation

• No pushback • Waste

– Unnecessary procedures • ½ of health expenses in last year of life

– Marginally useful procedures • Defensive medicine

– Bad Medicine • IOM: 48-98K “unnecessary” deaths/year

ObamaCare• Universal coverage: everyone must get insurance

– Employer – Insurance collaborative – Government (?) — rejected

• Insurance companies cannot deny insurance, cancel coverage, impose reimbursement limits based on illness, past or present

• Government assistance to poor people, small companies • Health Information Technology (HIT) to smooth info flow • Cost savings from avoiding billing disputes, ceasing to

reimburse only procedures, evidence-based medicine • Accountable Care Organizations

Quality Improvement

• IOM Study: 96,000 US deaths/year from medical error (perhaps half preventable?)

• Information intervention at the point of decision making can improve decisions

• DPOE: Direct Physician Order Entry allows such intervention

• Leapfrog Group: Large employers ($$$) require DPOE from providers

• Patient Involvement: Indivo Health, Google Health, Microsoft Healthvault

Implications of Health Care Organization for Informatics

• Money determines much – Historically, medicine spends 1-2% on IT, vs. 6-7% for

business overall, vs. 10-12% for banking – “Bottom line” rules, therefore emphasis on

• Billing • Cost control • Quality control, especially if demonstrable cost savings • Retention and satisfaction (maybe)

– Management by accountants

Challenges

• Computerized Medical Records (EMR/CPR/…)

• Usability of systems in the workflow of health care

• Large-scale “Engineering Systems” problem

• Genomic Medicine