NavHealth Dedicated Advisor Assignment

by Mark Silverberg ([email protected])

The Ask

You've been asked by a member ACO to solve the problem of finding the right providers to grow the network. The ACO has limited the scope to growing out the Cardiology network in particular. The attached 100 NPIs represent a nearby network of providers. Please use open data to learn what you can about these providers and make a recommendation (part of this exercise is your judgment on what data is relevant in selecting providers to partner with). Please pull together a short presentation (2-3 slides) on your key insights about the providers in question.

Note: The member is an Accountable Care Organization engaged in the Medicare Shared Savings Program.

Synthesis of the Ask

• Given a list of 100 NPIs and available open data, make a recomendation of which cardiology providers the MSSP ACO should recruit

• Rating will include judgement on which data is relevant (what contributes to the thanking)

• Deliverable: code/data and a 2-3 slide presentation

  • slide 1: restate the question, show funnel, and final rankings

  • slide 2: profile the two top cardiologists

  • slide 3: data sources and assumptions; rationale and ideas for next time

Work Completed

Notebooks

1. Filtering for only legal cardiologists

   • deduplicates the given 100 NPIs (down to 97)

   • join in demographic information from data source #1 (data.medicare.gov physician compare)

   • filter out non-cardiologists

   • check to make sure none of the cardiologists are on the HHS OIG Exclusions list

2. Digging into participation, utilization, cost, and quality

   • looks at which CMS programs the cardiologist participates in

   • brings in data from CMS Aggregate Utilization and Cost data for CY2014

   • starts to compare the 8 given cardiologists to the population of 5,480 CA cardiologists we have Medicare data on

3. Location-based analysis

Data Sources

1. data.medicare.gov "National Downloadable File" (NPPES) for demographic information such as name, specialties, and participation in certain programs

2. oig.hhs.gov "List of Excluded Individuals and Entities (LEIE)" for compliance ("Anyone who hires an individual or entity on the LEIE may be subject to civil monetary penalties.")

3. data.cms.gov "Medicare Physician and Other Supplier National Provider Identifier (NPI) Aggregate Report, CY 2014" for "information on utilization, payments (allowed amount, payment, etc.), and submitted charges organized by NPI"

4. Google Maps API via pygeocoder for geocoding provider addresses to lat/long coordinates so we can map them for geo analysis

Assumptions made

• The universe of providers (all_providers) is provided by the data.medicare.gov National Downloadable File

• the duplicates in this dataset are unimportant - we are taking only the first occurence of the NPI

• a NULL primary specialty indicates their specialty is unknown and we are not going to consider them as a potential provider in the ACO

• due to the lack of medical vs drug service data in source #3, only total service counts and costs were analyzed

• The customer is looking for only cardiologists and not cardiac surgeons

• population demographic distributions (age, gender, ethnicity, risk profiles) from source #3 which is from Medicare is consistent with reality for non-Medicare patients (doubtful, but it's the only data we have)

Research

Q: is it better to recruit providers whose patients have a higher or lower risk at cardiology-related conditions?

What makes a cardiologist attractive to an ACO?

• a focus on prevention (not just treatment)

• geographical diversity

What could we possibly use as proxies form open data to rank the providers in our given list?

• +1 for participation in million hearts campaign likely a good thing (none of our given providers are though 😦)

• +1 for PQRS and EHR participation (6/8)

• +1 for cardiac surgery secondary specialty (1/8)

• -1 de-prioritize cardiologists who are serving high-risk demographics? (data #3)

  • age?

  • gender?

  • race/ethnicity?

  • comorbid conditions? (blood pressure, diabetes, depression)

• +1 for volume of patients

• 7 of the 8 cardios in the dataset are male - should we prioritize a mix of genders?

  • borderline controversial/sexist perhaps. will make this a tie-breaker

• -1 for high risk (HCC) - lower chance of prevention working?

• geographic diversity - is not near other doctors, for gaining market share and being introduced to new communities/networks of people

Misc research tidbits

• Cardiologists are a great focus areas for ACOs

  • “Cardiology fits better than any other specialty because we are the most integrated specialty ... fit seamlessly into the ACO model because they already have invested in the skills and infrastructure at the foundation of ACOs. The specialty is data driven and has embraced quality metrics and electronic records, he said. In addition, 70 percent of cardiologists in the U.S. are already integrated." cardiovascularbusiness.com article

• Who is at risk for heart disease? NIH NHLBI

  • obviously, the more risk factors you have, the more at risk you are.

  • "More than 75 percent of women aged 40 to 60 have one or more risk factors for CHD"

  • risk factors include smoking, high blood pressure, diabetes, obesity, stress/depression, anemia, sleep apnea

  • "Risk Factors You Can't Control" include age (older age = higher risk)

Etcetera

Ideas

• Despite being given a list of 100 NPIs, only 8 ended up being cardiologists. We could look at the geographic region of the 100 NPIs and expand the search for cardiologists in the NPI master file. For exampler, perhaps an emphasis should be put on cardiologists participating in the Million Hearts program

• Productize this:

  • allow the upload of a list of NPIs which produces similar analysis

  • allow for interactive toggling of +/- n factors

Next time I would...