Correlation between Low Readmissions and High Death Rates

Heidenreich et al. pubmed.gov/20650356
(2010) Journal of the American College of Cardiology, 56(5), 362-368. 

• In a study of 50,000 patients at Veterans Affairs hospitals over 5 years, "Heart failure admission rates remained constant at about 5 per 1,000 veterans. Mortality at 30 days decreased (7.1% to 5.0%, p < 0.0001), whereas rehospitalization for heart failure at 30 days increased (5.6% to 6.1%, p = 0.11)" italics added

Gorodeski, et al. pubmed.gov/20647209
(2010) New England Journal of Medicine, 363(3), 297-298.

• "A higher occurrence of readmissions after index admissions for heart failure was associated with lower risk-adjusted 30-day mortality."

American Hospital Association aha.org/research/reports/tw/11sep-tw-readmissions.pdf
(2011) Trendwatch September 2011

• "analysis using Hospital Compare data conducted by the Greater New York Hospital Association also concluded that mortality is inversely related to readmissions. (Chart 3)"

• Chart shows that states with lowest mortality, MA, CT, DC, DE, MN, NJ, IL, OH, MI, PA, all have above average readmissions, and all but two of these states are in the 70th percentile of readmissions or higher.

Krumholz et al. pubmed.gov/23403683
(2013) Journal of the American Medical Association. 2013 Feb.13; 309(6): 587–593.

• They find 17% correlation between lower readmissions and higher deaths among heart failure patients. These are the same Yale authors who develop Medicare's official readmission data:
• "The analyses included ... 4767 hospitals for HF [heart failure] ... The correlations ... [of mortality and readmission rates] were ...−0.17" for heart failure.

Gilman et al. pubmed.gov/25092831
(2014) Health Affairs, 33, no.8 (2014):1314-1322

• "safety-net hospitals were more likely than other hospitals to be penalized under the... Hospital Readmissions Reduction Program... 
• "[M]ortality outcomes in safety-net hospitals were better than those in other hospitals for patients with acute myocardial infarction, heart failure, or pneumonia. 
• "Third, the adjusted cost per Medicare discharge was virtually identical at safety-net and non-safety-net hospitals. 
• "Taken together, these results indicate that safety-net hospitals provided better health outcomes than other hospitals at a similar cost level yet were more likely to be penalized under programs that are intended to improve and reward high performance."
  

Deaths Caused by a Program to Avoid Readmissions:

Fan et al. pubmed.gov/22586006
(2012) Annals of Internal Medicine 2012 May 15; 156(10):673-83

• 20 Veterans Affairs hospitals measured a "comprehensive care management program" for emphysema and chronic bronchitis (COPD). "Improving a patient's ability to self-monitor and manage changes"
• "data monitoring committee terminated the intervention before the trial's planned completion... There were 28 deaths from all causes in the intervention group versus 10 in the usual care group"

Minorities

Joynt et al. pubmed.gov/21325183
(2011) Journal of the American Medical Association. 2011 Feb 16;305(7):675-81

• "black patients were more likely to be readmitted after hospitalization"

Rodriguez et al. pubmed.gov/21835285
(2011) American Heart Journal. 2011 Aug;162(2):254-261.e3

• "Elderly Hispanic patients are more likely to be readmitted for HF and AMI [heart failure and heart attack] than whites"

Joynt et al. pubmed.gov/23340629
(2013) Journal of the American Medical Association. 2013 Jan 23;309(4):342-3

• "We found that large hospitals, teaching hospitals, and SNHs [safety net hospitals] are more likely to receive payment cuts under the HRRP [readmissions penalties]. It is unclear exactly why these hospitals have higher readmission rates than their smaller, nonteaching, non-SNH counterparts, but prior research suggests that differences between hospitals are likely related to both case mix (medical complexity) and socioeconomic mix of the patient population.2-3 There is less evidence that differences in readmissions are related to measured hospital quality.6"

Interviews

Dr Ashish Jha, of Harvard's School of Public Health, told PBS, "If you look at, for instance, the U.S. News [and World Report] publishes its list of top 50 hospitals. Those hospitals tend to have very low infection rates, very low mortality rates, very low death rates. Guess what? They tend to have very high readmission rates, because they do such a good job of keeping their patients alive that many of them are readmitted."

Dr. Sunil Kripalani, of Vanderbilt University Medical Center told Fox News, "Among patients with heart failure, hospitals that have higher readmission rates actually have lower mortality rates. So, which would we rather have -- a hospital readmission or a death?"

Dr. David Belk has clear data on wholesale (NADAC) and retail costs of generic and branded drugs (from GoodRx) and what drives the costs.

Drug stores in half the states are not allowed to tell you if the cash price is less than your co-pay, but the other half of states have laws letting them tell you. You can always ask anyway.

You can find the number of drug prescriptions from each doctor and costs for Medicare patients in at least 2 places, described below. For non-Medicare prices see above.

ProPublica has Medicare Part D cost for each drug: number of prescriptions and total spending. You can get separate totals for US and each state, so you can get average cost per prescription, and for each doctor who prescribed a drug 50 or more times in 2013.

Medicare itself has more complete Part D data. The US and state summary files (bottom of the link) show for each drug: the number of beneficiaries as well as prescriptions and spending, so you can get average per beneficiary (total during a year), as well as per prescription for each drug.

Medicare's detailed files show number of days prescribed, so you can get average cost of a daily dose, as well as each doctor who prescribed a drug 11 or more times in 2013. This info is in 23 million records, without state or US summaries. However you can get good state and national estimates by opening any of their 36 spreadsheets (divided by last name of prescriber) and getting averages there. The average costs do not vary much by last name of prescriber. (Tips for working with large spreadsheets)

Some doctors and drugs typically have 30-day or 90-day prescriptions, which may be renewed all year. Their averages include the cost for each  whole long prescription (30 or 90 days). Docs & drugs with shorter prescriptions only include that lower cost. Medicare's focus seems to be on cutting total costs, not cost per dose.

ProPublica's methodology says it has "retail cost" for these prescriptions.

Medicare's fact sheet gives more detail, saying it includes,

• The total drug cost includes the ingredient cost of the medication, dispensing fees, sales tax, and any applicable administration fees. It?s based on the amounts paid by the Part D plan, Medicare beneficiary, other government subsidies, and any other third-party payers (such as employers and liability insurers). Total drug costs do not reflect any manufacturer rebates paid to Part D plan sponsors.

It does not directly include patients' monthly premiums, though on average those premiums may cover all drugs, administration, and profits.

After seeing which drugs a doctor prescribes, you can find drug safety and effectiveness from the sources at the top of the page.

Most doctors send prescriptions to pharmacies electronically. This is more reliable than hand-written faxes, but two problems stand out:

• Most doctors cannot send, and most pharmacies cannot receive, electronic cancellation orders
• If the patient does not get a copy, the patient cannot check if the pharmacy filled it correctly.

Cancellation orders are crucial to correct mistakes and cancel refills. Patients can overdose when they keep getting the old medicine after the doctor orders a new one. Only a third of prescribers and 40% of pharmacies use software certified to handle cancellations, so less than a third of cancellations can go through. "Electronic health records allow prescribers to stop a prescription, but what many physicians may not realize is that in most cases that directive is not sent to any pharmacy," even though original prescriptions are reliably sent.

Health systems like the Veterans Health Administration and Kaiser Permanente, where prescriber and pharmacy are in the same organization, can cancel electronically. Otherwise only 5 pharmacy chains accept e-cancellations: 

• CVS/Caremark (which handles 23% of US prescriptions),
• ExpressScripts (12%),
• RiteAid (6%), 
• Kroger (3%),
• Wegmans (under 1%). 

Many non-chain pharmacies accept e-cancellation if they use one of 14 software systems listed with a † You can ask the pharmacy which software they use.

No other chain is certified for e-cancellation, such as Walgreens (16% of US prescriptions) and Walmart (7%). Consumer Reports unfortunately recommends Costco and Sam's Club for price (uncertified), Walgreens for its apps (uncertified), and supermarket chains for convenience (only Kroger's is certified). CR should know better.

Doctors can cancel electronically only if their software is certified under "Cancel Rx". Out of 920 systems, only 80 are marked as certified, and usually only the latest updates. If your doctor is part of a large group, you can ask the group to ensure its software gets certified and updated. Individual doctors have little control.

Automatic refills are even more dangerous. When pharmacies call patients to say, "Your prescription is ready," patients and callers do not know whether the doctor recently ordered it or it is a zombie renewal. Costco, CVS, RiteAid and Walgreens encourage patients to sign up for automatic refills, so patients at Costco and Walgreens (two which lack e-cancellations) can get undesired medicine for long periods, thinking their doctor ordered it.

Prior authorization for prescriptions is a system where an insurer tells a pharmacy that a doctor needs to send the insurer detailed information and get the insurer's approval for the prescription. Insurers tell pharmacies, not doctors or patients, when prior authorization is needed, and pharmacies say they have no obligation to tell doctors, though they have the forms and information which the doctor needs, and no one else does. A Massachusetts court says pharmacies must tell doctors, and a doctor says the entire circuitous system of prior authorization for drugs kills patients.

Comfort care's main justification is to provide a more enjoyable life for seriously ill patients than a life of treatment. However advocates also say it can extend life and save money. The sections below address these ideas.

There seem to be only 2 short, small, random studies of palliative care's effect on extending life. Both have almost complete overlap between the confidence intervals on length of life in palliative and control groups, which suggests little significant difference in survival.

The bigger study finds shorter life for palliative patients, and is ambiguous about its significance. The smaller, narrower study, of one disease in one hospital, claims significantly longer life for palliative patients, and only a bio-statistician can evaluate that claim.  Both studies are described below.

Medicare is pressing hospitals to limit care and promote hospice, comfort care (symptom relief or palliative care), and "do not resuscitate" (DNR) orders, so patients die at home and do not come back to the hospital. 

• In 2009 Medicare endorsed "end-of-life/palliative care programs" to cut costs and increase bonuses to doctors and hospitals.
• In 2012 a Congressional agency, MedPAC, recommended "hospice use and the presence of advance directives" to reduce rehospitalizations.

Huffington Post found some hospices earning money by enrolling patients who were not ill enough.

Mercy Health Center, OK (2009). Kroch et al, "Making hospital mortality measurement more meaningful: incorporating advance directives and palliative care designations." Am J Med Qual. 2010 Jan-Feb;25(1):24-33. pubmed.gov/19966112
"Patients with care-limiting orders have higher mortality than the general in-patient population; nevertheless most DNR patients (65%) still survive the hospital stay, albeit most PC [palliative care] patients (73%) do not ... Observed mortality rates for DNR and PC patients are generally higher than those expected from patient risk factors" (p.28)

Mercy Health Center, OK (2009). Kroch et al, "Making hospital mortality measurement more meaningful: incorporating advance directives and palliative care designations." Am J Med Qual. 2010 Jan-Feb;25(1):24-33. pubmed.gov/19966112
9,100 patients from 2005-2006, analyzing hospital records retrospectively. Included 995 with DNR, of which 311 had Palliative care.
Eligible: Hospital discharges and deaths Nov.'05-Oct.'06
Location: Mercy Health Center (hospital), Oklahoma City
Findings: Death rates during the current hospital stay were abnormally high for patients who had palliative care and/or DNR orders:

Hospital    Expected...  ...(Expectation Based on Patients' Condition)
  Death        Death
   Rate          Rate
   73%           31%       Palliative care patients
   35%           16%       DNR patients

This correlational study does not show if:

1. Lack of curative treatment caused the death rates to be so much higher than expected (expected rates take into account the history of the disease and other conditions, see below)
2. Counseling and acceptance of the end (or giving up) caused the death rates to be so high
3. These palliative and DNR patients would have died sooner without those orders, or later
4. There were any reasons why so few palliative patients went home to die

 

• Mercy's other patients had lower death rates than expected (p.25, they gave no figures)
• Most palliative care patients did not get to die at home (73%)
• This hospital considers both DNR and palliative care as "care-limiting" orders (p.24). 
• They frequently assign patients to palliative care whose disease is irreversible and leading to death, though death is not necessarily expected during the hospital stay, and "maintaining the patient's comfort during the dying process is the primary objective" (p.25) 
• "[G]eneral hypothesis that DNR designation identifies otherwise unobserved risk that is revealed over time during the hospital episode" (p.30) 
• Expected death rates are calculated by the CareScience method which controls for "age, sex, race, income, relative distance traveled, principal diagnosis, comorbidity-adjusted complication risk score, defining diagnosis, cancer status, chronic disease and disease history, valid procedures, admission source, admission type, payer class, and facility type" in each of "142 different disease groupings (ie. 142 distinct regression equations)" (p.26)
• The CareScience method is fairly good at predicting which patients will die during a particular hospital stay. It explains 30% to 54% of the variation in death rates (p.30)

National Hospice and Palliative Care Organization, US (2007). Connor et al, "Comparing hospice and nonhospice patient survival among patients who die within a three-year window." (J Pain+Symptom Manage. 2007 Mar;33(3):238-46. pubmed.gov/17349493
4,493 patients from 1998-2002, analyzing Medicare records retrospectively
Eligible: Patients with CHF and cancers of lung, pancreas, prostate, colon and breast who had a major progression in their disease ("indicative date") in 1999, but not 1998, who died between 15 days and 3 years after that major progression. There are detailed criteria for each disease (see below).
Location: US patients with Medicare Part B (optional doctor coverage)
Findings: Longer life for hospice patients with lung cancer or heart failure. This correlational study does not show if:

1. Longer-lived patients had more time to think about and sign up for hospice, or
2. Giving up curative treatment helped patients live longer,  or
3. Hospice's better coordination of care helped patients live longer, or
4. Other factors caused both longer lives and more hospice, such as better doctors and hospitals, health knowledge, other health experience, etc.

Significant Differences in Survival (days):

• Lung cancer 279 vs. 240, P<.00001
• Pancreatic cancer 210 vs. 189 P=.0102

Not Significant (over .05):

• Heart failure 402 vs. 321, P=.0540
• Colon cancer 414 vs. 381 P=.0792
• Breast cancer 422 vs. 410 P=.6136
• Prostate cancer 514 vs. 510 P=.8266

Potential Biases:

1. They omitted patients whose immediate treatment or lack of it led to death in the first 15 days. A majority of these omitted short lives could be patients who chose hospice, declined treatment, and died quickly, in accordance with their wishes.
2. They omitted patients whose treatment or hospice care carried them beyond 3 years. A majority of these omitted long lives could be non-hospice patients  with successful treatment.
3. They used detailed criteria for each disease, and they note their lung cancer criteria risked getting sicker patients into the control group than into the hospice group, since survival was measured from the last date in the records when lung patients switched chemotherapy drugs. Hospice patients stopped drugs, so stopped switching, while the control group kept switching, so their last switch date would be closer to death, and survival would be shorter in the control group. Breast cancer had the same bias, but still did not show a significant difference.

Healthgrades (2013) "Hospital Report Cards™ Mortality and Complications Outcomes 2013 Methodology"
"Top Five Risk Factors by Procedure or Diagnosis ... [p.45]

• Diabetic Acidosis and Coma ... DO NOT RESUSCITATE STATUS ... [p.46]
• Heart Failure ... DO NOT RESUSCITATE STATUS ... [p.47]
• Diagnosis code V49.86 (DO NOT RESUSCITATE STATUS) ... Healthgrades included this diagnosis as a risk factor in its regression analyses for all non-surgical cohorts where mortality was the outcome being assessed. It was only considered to be a risk factor when ... present on admission. This diagnosis was statistically significant in the logistic regression model for each of these cohorts." [p.50]

Hospice (even a strategic temporary signup) can provide more home care when needed, or reduce care that patients dislike. However it can also be pushed as a way to cut costs for the hospital or medical system:

American Academy of Hospice and Palliative Medicine (2011) "When patients enroll in hospice care, their days of hospitalization might be expected to go down, thanks to hospice’s 24-hour on-call capacity ... and the shift in goals of care."

DAI Palliative Care Group (2011) "Hospices and their palliative medicine specialists have proven, several studies have shown, to be effective at reducing use of hospitals for their patients... Hospitals will likely look to post-acute care networks to assist in managing the care of at-risk (for rehospitalization) patients. Should we consider deployment of palliative care specialists (physicians and nurse practitioners) by these networks to visit patients in their homes"?

New Jersey Hospital Association (2011) "Planning for 2012-2015..:

• Reductions in hospital readmission rates and penalties... More effective use of hospice
• Accountable care organizations and bundled payment... Home health and hospice have leading roles" (p.69)

Senior Housing News (2012) "Hospitals’ reimbursements will start getting docked under healthcare reform depending on 30-day readmission rates, so communities where many residents use hospice services rather than going to a hospital could be potentially benefit. [sic]"

National Quality Forum (2012) "For both Hospice and Palliative Care... treating ... symptoms ... has the strongest evidence base and helps avoid unwanted treatments and hospital/emergency department (ED) admissions and readmissions." (p.9)

Florida Hospital Association (2013) says one method they used to reduce readmissions 15% was "Evaluating the patient’s end-of-life care wishes" (p.8)

"Bon Secours [2014] already has reduced its readmission rates, improved palliative care for terminally ill patients..."

Southern California Public Radio (2014) " 'One of the major issues that we face is really trying to enhance end-of-life care,' Cedars-Sinai's Dr. Glenn Braunstein tells KPCC... He also noted that it's partially to help cut expenses, as the last month of someone's life in particular can be tremendously expensive."

Lawyers Dubler and Sabatino (1991) "system of allocation implemented inconspicuously by private institutions and practitioners. This sort of rationing will be difficult to uncover and even more difficult to prove and prevent. It will respond to implied regulatory messages from Medicare and Medicaid; it will react to reimbursement formulas and market reward. Most worrisome, it will couch personal and institutional prejudice in the language of medical and quasiscientific criteria." p. 116, "Age-based Rationing and the Law" chapter in Binstock et al, Too old for health care? : controversies in medicine, law, economics, and ethics, Johns Hopkins University Press

Jeffrey Birnbaum, (May 12, 1997), FORTUNE Magazine

• “AARP struggles to decide which of its members to put first: those who are already retired or the growing number who are still gainfully employed…
• has subtly begun to focus more on boomers by shifting the emphasis of AARP's publications and products to its nonretired members…
• Today its finances remain robust, anchored in health insurance as well as royalty-producing businesses that range from annuities to prescription drugs…
• would countenance other curtailments, primarily payment cuts to hospitals and physicians, in order to keep Medicare's hospital trust fund afloat…
• now AARP has competition on the left from the militant, five-million-plus-member National Committee to Preserve Social Security and Medicare and from the labor-backed National Council of Senior Citizens. On the right are newer, free-market groups like 60 Plus, United Seniors, and the Seniors Coalition, all small but growing…
• AARP must transform itself into something unusual--an oldsters' lobby that serves people of a wide variety of interests and ages… Some AARP executives even talk about renaming the magazine. After all, what young-at-heart boomer wants to read something called Modern Maturity?" [renamed AARP-The Magazine]
  

Alliance for Aging Research (2003) "Drawing upon scores of scientific studies, this important report shows how systemic bias against the elderly hurts older patients in the U.S.--highlighting ways in which the healthcare system fails older Americans. The report cites serious short-comings in medical training and prevention screening, and outlines treatment patterns that disadvantage older patients."

A 2007 study at Ohio State, Duke and other hospitals found that doctors and nurses "overestimate the risk of death," so they may limit care even among patients who have low risk of readmission and death.

Brown Professor Ackerman (2012) Doctors " 'keep a portion of the savings.' This arrangement obviously provides a financial incentive to withhold expensive life-prolonging treatment from Medicare patients whose quality of life is deemed low... Whose life is it, anyway? This slogan is conventionally used to support the right to die. It applies just as much to a sick old person who wants to stay alive. Such a person deserves better than to have well-schooled manipulators coax his family into signing his death warrant."

Neurologist Robert Weinmann (2012) "The days of searching out rare and unusual diseases to care for are over: these unfortunates will be obliged to find  whatever comfort is available under the nearest bus... Quietly, with as little fanfare as possible, physicians and hospitals will be encouraged to avoid the sickest, oldest, and most complicated patients."

Gastroenterologist Michael Kirsch (2013) "This is but a single example of how the medical profession is being forced to game the system to comply with a punitive financial penalty system that is poorly disguised as a medical quality initiative."

Cardiologist Walton-Shirley (2013) "palliative care has been birthed, ... of patients labeled as "frequent fliers" at high risk for budget-busting bounce-backs... a small part of me that worries that patients who need readmission will be held captive at home or in palliative-care programs or even become ensnared in a hospice-type situation when in fact there might have been help for them"

Medicare also researches how hospices spend Medicare's money: spending is higher early in a patient's participation in hospice, and just before death.

Starting in 2013, US hospitals are treating fewer patients for heart failure, and US death rates from heart failure are rising. Starting at the end of 2012, Medicare began penalizing hospitals for heart failure patients who were re-hospitalized (readmitted) within 30 days. Hospitals cannot always prevent readmissions, so the most effective way to avoid penalties has been to cut the number of Medicare patients they admit for heart failure.
 
Hospitals treated 60,000 fewer patients for heart failure in mid 2012-mid 2015, than in mid 2008-mid 2011, or 20,000 fewer patients during a year, compared to four years ago, before the penalties.
(Source, column CZ of: globe1234.org/hospitals1216.xls)

CDC says in the US:

• Death rates from heart failure fell every year from 2000 to 2012,
• Death rates from heart failure rose in 2013 and 2014.
• Source: cdc.gov/nchs/data/databriefs/db231.pdf, CDC instructions on defining causes of death: cdc.gov/nchs/data/misc/hb_cod.pdf

The higher death rates in 2013 and 2014 mean 7,200 and 9,600 more people died from heart failure in these years than would have died if the 2012 death rate had continued.
 
The term "heart failure" is also called "congestive heart failure" or cardiomyopathy. It refers to weak pumping because of muscle deterioration, stiffness, leaking valves, etc. It is not the same as a heart attack or heart stopping.
 
These are the latest national figures, but five earlier studies from 2010-2014 also found that hospitals which had fewer readmissions had more deaths, especially among heart failure patients.
(Source: globe1234.info/medicare/category/research)
 
Medicare said in August 2012, "We are committed to monitoring the measures and assessing unintended consequences over time, such as the inappropriate shifting of care, increased patient morbidity and mortality, and other negative unintended consequences for patients." (p.53376) They have not reported any of these monitoring results in 4 years.
(Source: federalregister.gov/d/2012-19079/p-1799)
 
The penalties apply to patients treated under Medicare Part B. Hospitals which face the readmission penalties now admit 5% fewer Part B patients for heart failure, even though the total number of seniors covered by Part B increased 12% in the same period.
 
Readmission penalties give hospitals incentives to treat fewer seniors. Medicare even gives hospitals an online tool to predict readmission risk for each potential patient.
 
Hospitals can avoid penalties by any mix of the following:

• Avoid admitting the sickest Medicare patients with heart failure ("There's not much we can do for you. Hospitals are dangerous. You're better off at home.")
  
• Treat as many as possible of the least sick outside of hospitals
  
• Change diagnosis to "hypervolemia," too much water in the blood, which is not penalized, but risky if caught
  
• Improve subsequent care for those admitted, to reduce readmissions
  

It is easier to give less care than to improve it, though hospitals certainly are doing both. And what we see is that death rates have started to rise.
 
The following hospitals had the biggest drops in heart failure patients admitted, comparing the most recent 3-year period to the 3 years before penalties:
 
St Vincent's Medical Center Riverside, Jacksonville, FL, -871 patients
Northwest Community Hospital 1, Arlington Heights, IL, -779 patients
Baptist Medical Center, San Antonio, TX, -724
Community Medical Center, Toms River, NJ, -570
St Luke's Hospital Bethlehem, PA, -543
King's Daughters' Medical Center, Ashland, KY, -536
Beaumont Hospital - Dearborn, MI, -517
Hackensack University Medical Center, NJ, -504
Vassar Brothers Medical Center, Poughkeepsie, NY, -454 patients
 
On the other hand these hospitals may have unique reasons for their changes, and the real story may be among all the other hospitals with smaller drops in heart failure patients. Changes at all hospitals are in a spreadsheet (in column CZ; changes in Part B enrollment are in column DL):
globe1234.org/hospitals1216.xls
 
I counted hospital admissions in July 2012-June 2015, compared to July 2008-June 2011. These are the newest and oldest comparable data available. Medicare released the older data in a comparable form in May 2013. It released the newer data in August 2016.

Hospitals face readmission penalties when they treat Medicare patients for heart failure. Each hospital pays a penalty if more of their heart failure patients than the US average need another hospital stay within a month. So hospitals know they have a 50% chance of a penalty, since about half the hospitals will have readmission rates above average each year. 

Meanwhile for heart attacks, admissions fell 1.3% in the same time period, even though Part B beneficiaries increased 12%. I haven't found death rates from heart attacks, and it seems that all survivors who reach a hospital would be admitted. Have heart attack rates really dropped 13%? Are more people dying before they reach a hospital? Or is something else driving down hospital admissions for heart attacks?
 
For pneumonia, which is the other of the three original readmission penalties, we have to compare 3 year periods ending June 2014 and June 2011, since Medicare expanded the pneumonia categories counted in later periods. Pneumonia admissions fell 4% over that period, while the number of Part B beneficiaries rose 9%. Death rates oscillate each year but were on a downward trend from 1999-2012. It looks as if the trend may not have continued in 2013 and 2014, though it is hard to tell.
statista.com/statistics/184574/deaths-by-influenza-and-pneumonia-in-the-us-since-1950
cdc.gov/nchs/data/health_policy/influenza-and-pneumonia-deaths-2008-2015.pdf
 
Readmission penalties are large. Hospitals get $6,000 for treating a Medicare heart failure patient, but pay a $27,000 penalty for each readmission within 30 days, above the national average rate. For other conditions penalties range from $25,000 to $239,000 per readmission above the national average rate. So every hospital tries to be below the average, driving the average down and the risk of penalties up every year. There are also minimal adjustments for the mix of patients each hospital serves. Penalties total $469 million this year.

There are newer penalties for re-hospitalizing patients after coronary bypasses. The penalty is $188,000 for each one above the national average rate; penalties began October 2017. Penalties after elective hip and knee replacements are $239,000 and began October 2014. The penalty calculations are written into the Affordable Care Act. It is too early to see if the number of people treated has fallen, but the  American College of Surgeons warned Medicare that treatment would be cut: "the potential that these hospitals will decrease their care for such patients, thereby creating an access issue."
(Source: regulations.gov/contentStreamer?documentId=CMS-2013-0084-0090&attachmentNumber=1&disposition=attachment&contentType=pdf)
  
In 67 metro areas, Medicare has a second way to discourage hip and knee replacements, especially for the frailest patients who may need them most: the hospital must pay nearly all medical expenses for 90 days after the hospital stay, though it has no control over these costs. Fewer hip and knee replacements and fewer coronary bypasses, when Medicare patients need them, condemn seniors to reduced activity and faster decline.
(Source: globe1234.info/medicare/publiccomment)
 
For heart attacks and coronary bypasses, Medicare plans the same approach of making hospitals pay 90 days of medical costs, starting July 2017, in 98 metro areas.
federalregister.gov/d/2016-17733/p-3
federalregister.gov/d/2016-17733/p-753

Another page explains some arithmetic behind the readmission penalty calculations, which give hospitals a strong incentive to serve fewer patients.