open404 (IT Remix)

data intermission

important part...just a start

?

Do the units per case for ACT match the frequency required for fidelity?

What does it mean when CLS Daily and Personal Care diverge or converge?

Are any CMHs leading in the move to decrease residential services and move people into community-based settings?

Why would a CMH serve more people in ACT than Case Management and then reduce the number of cases in each to < 1/4 of the original number?

How much do provider admin costs impact the

unit cost of services?

Does a steady decrease in health services correspond to a greater use of non-Medicaid health services?

One CMH provided CLS Daily to one less person in 2012 than in 2006 but spent $2M more doing it. Why?

a.k.a. Hypothesis generation

Any Questions?

Who has more incentive to make sure the data is right than the person the data describes?

• Inconsistent CPT/HCPCS code use
• Ambiguous variable definition
• Difficult to make meaningful comparisons of population, service use, and gaps
• Little incentive to use it as a tool

• Compiled into master dataset
• New variables: Fiscal Year, PIHP, % cost, % use, Service Types and sub-groups
• Standardized units to hours
• Aggregated by population and region
• Reformatted to look at correlations

Caveats:

"but the data isn't right..."

Some starting

*

Resource Use / Efficiency

Utilization Management

ideas

Look at instances where:

• different service codes are provided in varying amounts within a given service type
• year-over-year use of high-cost services is flat or increasing

Hone in on areas where...

• high cost services are provided often to a small group
• high variation exists across CMHSPs for per-person use of a service
• high frequency services are provided with high units/person

...opened up

Use it.

• Required by the state legislature
• Compiled by MDCH BHDDA
• Contains service cost and utilization
• Aggregated by CMHSP, disability type
• Distributed as PDF
• Has potential…

Greatest

obstacle

to data-

driven

decisions?

...as you know it

Our brains.

Screening for human:

17 x 24 = ?

1 million + 1 million = ?

Let's try it out...

So, did you really solve the first one?

assumed green = good for KPIs

focused on how data was presented rather than the data itself?

ignored evidence that contradicted your preconceived notions?

Have you ever...

believed a report because it told a coherent story, without asking about the quality of evidence?

one little thing...*

097654984186989085498756750976469761497654976

The 404...

open404

Strings attached

PDF text

Cost

Few

entails business risks, including reputational issues related to the potential release of negative information

Free

No limits

Everyone

CSV, JSON

& more...

Degrees of Open-ness

Cost

Rights

Access

Machine readable

What next?

Roll credits...

• Ask hard questions
• Prioritize new datasets to open
• Invest in tech & talent to collect & analyze
• Use it*

This presentation and tool would not have been possible without the work of…

• Catherine Idema
• Kathy Haines
• Kasi Hunziger
• Hope Network

Why would we want that?

Availability & Accessibility

Reuse & Redistribution

Open Data

Transparency

Innovation

Engagement

Value

What is it?

Data is useful.

Data is available.

It needs improvement…

…which only happens if we use it.

Adapted from Open Knowledge Foundation: http://okfn.org/opendata/

*Make it useful:

• Standardized data collection & reporting
• Clear variable definitions
• Agreed upon outcome measurements
• Easily accessible in machine-readable format

McKinsey Global Institute (2013) "Open data: Unlocking innovation and performance with liquid information"

• customize actions for consumer populations
• better match supply with demand
• allow innovative approaches by market entrants

[plug your ideas into a need]

Insight into:

• price/services for consumers
• variability

"the use of open and proprietary data in health care could help generate value of $300 billion to $450 billion per year in the United States"

Potential benefits include:

• increased efficiency
• new products and services
• better services

Innovation

Engagement

- Justice Louis Brandeis

Value

"Sunlight is

disinfectants"

...the best of

Transparency

knowledge is power

sharing

• Version control (distributed)
• Command line = bash shell a la Linux
• Stick a Fork in it
• Clone it on your desktop

• Integrated development environment for R
• Create Markdown, Sweave documents for reproducible analysis
• Projects integrated with Git/SVN for version control

Open gifts

open404

wake

clean & analyze

visualize

https://github.com/j-hagedorn/open404

ask

more data

> than #s

RStudio :: http://www.rstudio.com/ide/download/

path

RAW :: http://raw.densitydesign.org/

ManyEyes :: http://www-958.ibm.com/software/data/cognos/manyeyes/

zoom

SAMHDA :: http://www.icpsr.umich.edu/icpsrweb/SAMHDA/browse

BRFSS :: http://www.cdc.gov/brfss/annual_data/annual_2012.html

Medicare Provider Charge Data :: http://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/Medicare-Provider-Charge-Data/Physician-and-Other-Supplier.html

tell

Purpose

Like data?

Teach to fish

Give access to data

Provide free tools

Yes

No

Give fish