Module 3: Medicare Advantage Quality and Regression Discontinuity

class: center, middle, inverse, title-slide

# Module 3: Medicare Advantage Quality and Regression Discontinuity
## Part 1: MA Data and Quality Measurement
### Ian McCarthy | Emory University
### Econ 470 & HLTH 470

---

<style type="text/css">
.remark-slide-content {
    font-size: 30px;
    padding: 1em 2em 1em 2em;    
}
.remark-code {
  font-size: 15px;
}
.remark-inline-code { 
    font-size: 20px;
}
</style>

# What is Medicare Advantage
- Private provision of health insurance for Medicare beneficiaries
- Medicare "replacement" plans <br>

--
- It's just private insurance for those eligible for Medicare

---
# Medicare Advantage History
- Existed since 1980s, formalized in the 1990s, expanded in 2000s
- Medicare+Choice as part of Balanced Budget Act in 1997
- Largest expansion: Medicare Modernization Act in 2003 (also brought Medicare Part D)

---
# Medicare Advantage Details
In its current form...
- Insurers submit plan details and a price needed to cover traditional Medicare ("bid")
- If approved, Medicare pays risk-adjusted bid *or* benchmark
- Bid `\(<\)` benchmark, insurer gets a rebate
- Bid `\(>\)` benchmark, insurer charges premium
- Seperate bidding for Part D

---
# Medicare Advantage in Real Life
Let's take a look at the Medicare Advantage plan options...

--
<br>
[Medicare Plan Finder](https://www.medicare.gov/plan-compare/)

---
# Medicare Advantage Quality Ratings
- Initial MA Star Ratings (2007)
- Overall rating introduced in 2009
- Complicated formula...

- **key point:** ratings from several domains are averaged and then rounded
- More details available [here](https://www.cms.gov/Medicare/Prescription-Drug-Coverage/PrescriptionDrugCovGenIn/PerformanceData)

---
# Role of Quality Ratings

- Potentially affect plan choice
- Quality bonus program in 2012
  - Demonstration program at first (2012 through 2015)
  - ACA quality bonus payments begin 2015
  
---
# Details of QBP

Eligible plans may receive two bonuses:

1. Benchmark bonus: Increased benchmark payment of up to 5\%
2. Rebate bonus: Share of bid-benchmark differential going back to insurer (from 50% up to 70%)

- Applied to plans with 3-stars and above from 2012-2015 (demonstration period)
- Applied to plans with 4-stars and above beginning 2015

---
# Some good primers

- [KFF Bonus Payment Illustration](pics/kff-illustration.pdf)
- [2021 Star Ratings and Bonus Payments](https://www.kff.org/medicare/issue-brief/medicare-advantage-in-2021-star-ratings-and-bonuses/)

---
# Medicare Advantage Data
- Recall the Medicare Advantage repository, [Medicare Advantage GitHub repository](https://github.com/imccart-test/ma)

--
- Now we need to work with the **final** dataset

---
# Final MA Data

```r
ma.data <- read_rds(here("data/final_ma_data.rds"))
```

---
# Summary stats
Focus on enrollments and star ratings:

```r
stargazer(as.data.frame(ma.data %>% select(avg_enrollment, avg_eligibles, Star_Rating)), type="html")
```

<table style="text-align:center"><tr><td colspan="8" style="border-bottom: 1px solid black"></td></tr><tr><td style="text-align:left">Statistic</td><td>N</td><td>Mean</td><td>St. Dev.</td><td>Min</td><td>Pctl(25)</td><td>Pctl(75)</td><td>Max</td></tr>
<tr><td colspan="8" style="border-bottom: 1px solid black"></td></tr><tr><td style="text-align:left">planid</td><td>895,495</td><td>35.849</td><td>69.789</td><td>1</td><td>4</td><td>41</td><td>999</td></tr>
<tr><td style="text-align:left">avg_enrollment</td><td>204,173</td><td>397.404</td><td>1,578.641</td><td>11.000</td><td>31.250</td><td>231.667</td><td>63,234.080</td></tr>
<tr><td style="text-align:left">avg_eligibles</td><td>749,267</td><td>42,587.920</td><td>98,741.840</td><td>11.571</td><td>3,798.000</td><td>37,116.570</td><td>1,355,734.000</td></tr>
<tr><td style="text-align:left">Star_Rating</td><td>448,793</td><td>3.307</td><td>0.783</td><td>1.500</td><td>2.500</td><td>4.000</td><td>5.000</td></tr>
<tr><td colspan="8" style="border-bottom: 1px solid black"></td></tr></table>

---
# Clean the data
Limit to plans with:
- Observed enrollments, `\(>\)` 10
- First year of star rating (2009)
- Observed star rating

--
<br>

```r
ma.data.clean <- ma.data %>%
* filter(!is.na(avg_enrollment) & year==2009 & !is.na(partc_score))
```

---
# Calculate raw average rating

```r
ma.data.clean <- ma.data.clean %>%
  mutate(raw_rating=rowMeans(
    cbind(breastcancer_screen,rectalcancer_screen,cv_cholscreen,diabetes_cholscreen,
          glaucoma_test,monitoring,flu_vaccine,pn_vaccine,physical_health,
          mental_health,osteo_test,physical_monitor,primaryaccess,
          hospital_followup,depression_followup,nodelays,carequickly,
          overallrating_care,overallrating_plan,calltime,
          doctor_communicate,customer_service,osteo_manage,
          diabetes_eye,diabetes_kidney,diabetes_bloodsugar,
          diabetes_chol,antidepressant,bloodpressure,ra_manage,
          copd_test,betablocker,bladder,falling,appeals_timely,
          appeals_review),
    na.rm=T)) %>%
  select(contractid, planid, fips, avg_enrollment, first_enrollment,
         last_enrollment, state, county, raw_rating, partc_score,
         avg_eligibles, avg_enrolled, premium_partc, risk_ab, Star_Rating,
         bid, avg_ffscost, ma_rate)
```

---
# Distribution of star ratings

```r
ma.data.clean %>% 
  ggplot(aes(x=as.factor(Star_Rating))) + 
  geom_bar() +
  labs(
    x="Star Rating",
    y="Count of Plans",
    title="Frequency Distribution of Star Ratings"
  ) + theme_bw()
```
.plot-callout[
<img src="03-ma-quality1_files/figure-html/rating-count-callout-1.png" style="display: block; margin: auto;" />
]

---
# Distribution of star ratings

---
# Enrollments and star ratings

```
## 
## Call:
## lm(formula = avg_enrollment ~ factor(Star_Rating), data = ma.data.clean)
## 
## Residuals:
##    Min     1Q Median     3Q    Max 
##   -627   -388   -214    -51  41908 
## 
## Coefficients:
##                        Estimate Std. Error t value Pr(>|t|)    
## (Intercept)               87.31      43.32   2.016  0.04387 *  
## factor(Star_Rating)2      32.75      50.62   0.647  0.51758    
## factor(Star_Rating)2.5   194.65      47.15   4.128 3.67e-05 ***
## factor(Star_Rating)3     433.95      49.84   8.707  < 2e-16 ***
## factor(Star_Rating)3.5   470.91      53.47   8.808  < 2e-16 ***
## factor(Star_Rating)4     552.30      57.91   9.538  < 2e-16 ***
## factor(Star_Rating)4.5   272.36      82.68   3.294  0.00099 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1440 on 17451 degrees of freedom
## Multiple R-squared:  0.01559,	Adjusted R-squared:  0.01526 
## F-statistic: 46.07 on 6 and 17451 DF,  p-value: < 2.2e-16
```