Rational Adjustment After a Chronic Diagnosis:
Evidence from New Type 1 DiabeticsFebruary 6, 2020 — Alex Hoagland
Outline1. Bio (5 mins)2. Research Presentation (35 mins)3. Grad School Life (15 mins)4. Other Q’s
Bio
About meUndergraduate: Brigham Young University
B.A., Economics & B.S., Mathematics (2017)
● Got into economics b/c of HS debate
● Empirical microeconomics
● Started out thinking of development
● Decided to do a PhD b/c I enjoy research
and its impacts
About meUndergraduate: Brigham Young University
B.A., Economics & B.S., Mathematics (2017)
● Got into economics b/c of HS debate
● Empirical microeconomics
● Started out thinking of development
● Decided to do a PhD b/c I enjoy research
and its impacts
Graduate School: Boston University
PhD, Economics (2021)
● Current third year (mostly research)
● Switched into health economics last year
● Work with Randy Ellis (BU Econ), Tal Gross
(Questrom)
Research(in progress)
My Research AgendaHealth Economics: how traditional economics/incentives are distorted in health care
— Moral hazard and adverse selection
— Payment systems and risk/insurance
— Provider, patient, physician behaviors
My Research AgendaHealth Economics: how traditional economics/incentives are distorted in health care
— Moral hazard and adverse selection
— Payment systems and risk/insurance
— Provider, patient, physician behaviors
● Innovation in Health Care: how scientific advancements are integrated
● Mental Health Care: an often overlooked (but increasing) part of care
● Behavioral Responses in Health Care: integrating behavioral economics
Documented Costs of Chronic IllnessChronic Illness: Illness lasting >1 year, requiring ongoing medical attention
● Health insurance becomes really important! But also more difficult to get:
● Pre-existing condition exclusions limit access to adequate care (Stroupe et al., 2000)
○ Even if coverage can’t be denied, benefits can be withheld for a certain time
● Increased need for health insurance leads to job lock
○ Employer mobility reduced as much as 40% (Beatty and Joffe 2006; Stroupe et al., 2010)
○ Prevents (i) finding a better job, (ii) internal advances, and (iii) negotiating power (e.g., retirement)
Healthcare Shopping● Little evidence that consumers shop around for care
○ Employers typically can’t evaluate the best choice of plan — too much information to sift through!
○ High degree of information friction (what is an HDHP?) and hassle cost (calculating expenditures)
to make rational decision attainable (Handel & Kolstad, 2015)
○ What’s more, choosing your doctor is also hard: prices, scheduling, proximity, etc..
○ Many people just default into whatever plan/provider is easiest for them.
Healthcare Shopping● Little evidence that consumers shop around for care (Handel 2013)
○ Employers typically can’t evaluate the best choice of plan — too much information to sift through!
○ High degree of information friction (what is an HDHP?) and hassle cost (calculating expenditures)
to make rational decision attainable (Handel & Kolstad, 2015)
○ What’s more, choosing your doctor is also hard: prices, scheduling, proximity, etc..
○ Many people just default into whatever plan/provider is easiest for them.
● Some attempts to address this:
○ High-deductible health plans (HDHPs): developed to give consumers “skin in the game”
○ Typically only reduce utilization without increasing shopping (Sinaiko et al., 2018)
● My question: What happens when skin comes before choices?
○ A chronic diagnosis (should) change how you approach health care.
○ You are more likely to benefit from generous coverage (less risk)
○ Your doctor-patient match matters more as well (chronic illness stigma, Gudzune et al., 2013)
How do people respond to chronic diagnoses?When you are diagnosed with a chronic illness, how do you change your:
1) Choice of health insurance plan?
2) Choice of primary care provider?
3) Overall health utilization?
4) Health spending throughout the year (deductible response)?
How do people respond to chronic diagnoses?When you are diagnosed with a chronic illness, how do you change your:
1) Choice of health insurance plan?
2) Choice of primary care provider?
3) Overall health utilization?
4) Health spending throughout the year (deductible response)?
○ A deductible is meant to deter/control adverse selection effects
○ But if you have a chronic illness, you’re not at risk (demand becomes inelastic)
○ How does this change decisions for things that are typically “post deductible”?
Specifically, do you switch?
Context: Type 1 DiabetesA chronic illness limiting an individual’s ability to produce insulin. Useful b/c:
1) Primarily affects children, so more likely to affect an employee’s dependents
2) No selection concerns: disease is not driven by health behaviors
3) Coverage is extremely important (high insulin prices, deductible typically met)
4) Treatment/management can require/benefit from durable medical equipment
purchases (insulin pumps, continuous glucose monitors, etc.)
Data: Truven MarketScan Claims DataA data set of all health claims from select companies/employers providing private
insurance to their employees. Includes information on:
● Enrollment (demographics, type of plan, industry, etc.)
● Inpatient/outpatient/pharmaceutical utilization
● Type of provider, diagnoses, procedure codes
● Payment info (how much was your copay? Deductible? etc.)
Covers 2007 — 2018 (BU has a subscription that keeps data up to date)
● For access to data, talk to Randall Ellis ([email protected])
My sample: About 40,000 individuals diagnosed with Type 1 Diabetes, + families
First Step: Identifying the Diabetics (Data Cleaning)There are two main ways that someone shows up as “diabetic” in the sample:
1) Diagnosis code (easy to screen out Type 1 from other types)
2) Procedure for an hemoglobin A1C lab test (not as straightforward)
There are about 1.8 million individuals who fall into this category.
First Step: Identifying the Diabetics (Data Cleaning)There are two main ways that someone shows up as “diabetic” in the sample:
1) Diagnosis code (easy to screen out Type 1 from other types)
2) Procedure for an hemoglobin A1C lab test (not as straightforward)
There are about 1.8 million individuals who fall into this category. To be sure I am
catching (i) type 1 diabetics (ii) at their time of diagnosis, I require:
● Enrollment for at least a year without taking insulins or any diabetes diagnoses
● Any elevated A1C’s to be confirmed by a physician diagnosis
● Follow-up visits or prescriptions for at least 6 months following initial diagnosis
Data Check: Visits to Endocrinologists
Question 1: Do Diagnosed Patients Change Plans?
Okay, what does this mean? A couple of funky things going on with this graph:
● Drop in plan switching at time of diagnosis, but
● Lots of other spikes (particularly +/- a year around dx time)
What concerns us about this graph?
Okay, what does this mean? A couple of funky things going on with this graph:
● Drop in plan switching at time of diagnosis, but
● Lots of other spikes (particularly +/- a year around dx time)
What concerns us about this graph?
● Maybe we’re seeing “phantom diagnoses” — people just happen to get treatment
as they come into the sample
● How many of these employees really have a choice in their plans?
● How does the type of insurance plan they start with affect this?
Okay, what does this mean? A couple of funky things going on with this graph:
● Drop in plan switching at time of diagnosis, but
● Lots of other spikes (particularly +/- a year around dx time)
A primer on health (for those who don’t immerse themselves in this jargon every day):
PPO: Good coverage w/ low deductibles. (higher premiums). Lots of provider choice.
HMO: Lower cost, more restricted network-style care.
HDHP: Same freedom as a PPO with lower premiums and much higher deductible
Newly diagnosed don’t want to lose premium coverage!
Responses are more muted from other plans
Responses are more muted from other plans
One more check: January diagnosesMany plans have open
enrollment during January.
Could the spikes be due to
well-timed diagnoses?
One more check: January diagnosesMany plans have open
enrollment during January.
Could the spikes be due to
well-timed diagnoses?
Response seems to be
unchanged if we ignore
January diagnoses!
Exploring Rational AdjustmentTwo kinds of behavioral biases may be at play here:
1) Anchoring Bias: Where you start matters (starting in PPO vs. HMO/HDHP)
a) Related issue is loss aversion — don’t want to lose the good you have
b) Is failing to switch out of a PPO evidence of shopping? Or something else?
Exploring Rational AdjustmentTwo kinds of behavioral biases may be at play here:
1) Anchoring Bias: Where you start matters (starting in PPO vs. HMO/HDHP)
a) Related issue is loss aversion — don’t want to lose the good you have
b) Is failing to switch out of a PPO evidence of shopping? Or something else?
2) Recency Effect: How close your diagnosis is to your plan choice may matter
a) If I have to choose a plan right after being diagnosed, maybe I do my homework
b) On the other hand, if I’ve had 10 months to get used to the new status quo, maybe I
don’t need to shop —> continue defaulting
Does the Behavior Change Save Money?
Does the Behavior Change Other Family Members’?
Does the Behavior Change Other Family Members’?
Digging Deeper: IdentificationWant to get at the causal impact of diagnosis on:
● Plan switching
● Provider switching
● Health utilization (particularly durable medical equipment spending)
How can I do that?
Digging Deeper: IdentificationWant to get at the causal impact of diagnosis on:
● Plan switching
● Provider switching
● Health utilization (particularly durable medical equipment spending)
How can I do that?
1. Exploit plausible randomness in diagnosis timing/onset
2. Identify a control group: propensity score matching
a. Identify similar individuals based on family structure, previous health spending, etc.
3. Then, can use event study design
a. Generalized difference in differences — what is the dynamic effect of a diagnosis?
Digging Deeper: To DoChronic illnesses change behaviors. How does a chronic diagnosis affect choices of:
● Health plans?
● Providers?
● Utilization?
● Investments in expensive health equipment?
Is there evidence of rational adjustments to new complications?
● Deductible Irrelevance
● Plan upgrading
How do these responses drive changes in utilization, costs, and specialist fees?
Grad School Life
Becoming a Grad Student● Why decide to go? Interested in research/teaching, gain quant skills, other data positions
● How to prepare?
○ Prioritize research opportunities, consider Econ/Math, pre-docs
● Who to talk to about going?
○ Faculty members, current grad students, people who left academia
● Application process:
○ Taking the GRE
○ Picking schools
○ Getting letters of recommendation
○ Completing the applications (letters of intent, writing samples, etc.)
● Choosing a school: Consider faculty, ambience (flyouts), less so about weather/stipend
Grad Student Life● Year 1: Standard classes (micro/macro/metrics) + qualifying exams
● Year 2: Field classes — generally get to try out 3 or 4 fields
○ Complete second year paper
● Year 3: Finish up courses, start your research pipeline
● Beyond: Need a job market paper, 3 chapter dissertation, RA work
● Typical length: 6 years (yes, that is forever)
● Other caveats:
○ Your specialty is likely to change — be open to new ideas/general ed!
○ Selection into grad school is very noisy signal — don’t put too much into it
○ Book recommendation: A Guide for the Young Economist, William Thomson
Other [email protected]