Does Sorry Work the Impact of Apology Laws on Medical Malpractice JRU final 1

ABSTRACT Physicians’ apologies for adverse medical events are acknowledged as a factor in patients’ decisions to litigate. Apology laws which render physicians’ apologies inadmissible in court are written to encourage patient-physician communication and to overcome the physicians’ disinclination to apologize because apologies could invite lawsuits. We present a novel model of apologies and malpractice in order to examine whether state-level apology laws have an impact on malpractice lawsuits and settlements. Using a difference-in-differences estimation, we find that apology laws could expedite the resolution process. We also find that apology laws result in the greatest reduction in average payment size and settlement time in cases involving severe patient outcomes. JEL Classification Codes: K13, K32 2 “It’s sad, so sad Why can’t we talk it over? Oh, it seems to me That sorry seems to be the hardest word” ~ “Sorry Seems to Be the Hardest Word”: Elton John Medical treatments often pose a probability that the patient will suffer adverse health effects. The probability of these adverse effects depend on the level of care exercised by the physician, which is not observable by the patient. Adverse outcomes can lead to medical malpractice claims. However, apologies by physicians can potentially provide a signal to the patient of the level of care, and consequently will affect the rate of litigation and medical malpractice costs. This paper presents a model of apologies and medical malpractice litigation and estimates the effect of state level laws regarding medical malpractice and apologies. In response to trends in malpractice lawsuits and malpractice insurance premiums, national and state legislatures have proposed and enacted legislation to put into place a number of reforms, including jury award caps, insurance premium price caps, state medical malpractice funds, greater information disclosure, and of particular relevance to the current paper, apology exemptions. In 1986, Massachusetts became the first state to adopt some form of an apology law, which was designed to protect doctors from statements they might make to their patients regarding complications the patient suffered under their care. More recently, these apology laws specifically declare that a statement of apology made by a medical practitioner to a patient is inadmissible as evidence of liability. In September of 2005, then-Senators Clinton and Obama cosponsored the National MEDiC Act, which would have implemented such exemptions on a national level. As of January 2009, 36 states had implemented various forms of apology laws. State apology laws are premised on two stylized facts. The first fact is that doctors would like to apologize to their patients for medical mistakes, but are stymied by their fear of inviting a lawsuit. Research shows that doctors are typically told to avoid admissions of fault and apologies because of the 3 risk of lawsuits (Lamb et al., 2003). The second fact is that a main motivation patients give when asked why they chose to sue their doctors is anger; an anger that might have been assuaged by an apology (Vincent, Phillips & Young , 1994). These two facts lead to a vicious cycle that breaks down patientdoctor communication and thereby increases litigation costs. By making apologies inadmissible as evidence in malpractice lawsuits, it is hoped that doctors would communicate with patients more effectively, thus reducing patient confusion and anger in the case of adverse complications and thereby diminish the cost of medical malpractice by eventually reducing malpractice insurance losses and premiums. Our study examines the impact of state-level apology legislation. Although we do not observe actual apologies, the maintained assumption of this paper is that by reducing the consequences of apologies, doctors would apologize more frequently. The purpose of apologies is to maintain the doctorpatient relationships. Our model presumes that patients care about their doctors’ intentions, but the relationship between a patient’s health outcomes and a doctor’s intentions is fundamentally uncertain. Patients are more inclined to sue a doctor with poor intention, (i.e., doctors who put in effort below the standard of care). An apology is a signal that the bad outcome arose from bad luck rather than a lack of effort.1 More apologies would therefore reduce litigation and speed up settlements. However, the law could have perverse consequences. By reducing the cost of the apology, the laws could make apologies less effective and therefore lawsuits may become more likely. Another unintended consequence may be that patients may be induced to litigate more frequently if more apologies make patients better informed about the extent and the cause of their injuries. To disentangle these effects we use data from the National Practitioner’s Data Bank (NPDB) Public Access File (2009), which includes all medical malpractice payments made by or on behalf of a 1 A more extensive model of apologies based on signaling and asymmetric information can be found in Ho (2011). In order to keep the focus on the litigation and maintain simplicity in the current article, we abstract away from signaling by saying that an apology increases the psychic cost of suing a doctor, implicitly assuming the patients’ beliefs change. 4 healthcare provider since 1991. Given that the NPDB data set only consists of claims with positive payouts, it does not contain information on open claims nor closed claims without payments. Therefore, the dependent variable in most analysis would be the number of closed claims with positive payouts in a state-year. Our analysis is conducted on two levels. The first level is a difference-in-differences analysis using the time variation from the staggered implementation of state-level apology laws to investigate the impact on the malpractice closed claims frequency and malpractice compensation payout within a stateyear. We further explore the change of claim composition by the severity of injury. The second level of analysis is a duration analysis on the impact of apology laws on time to resolution (i.e., how long it takes for a malpractice suit to be resolved), which is conducted at the individual level. We further investigate the impact of apology laws on claim payment and how the size of this impact depends on the severity of medical injury. The results from difference-in-difference would be biased if the passage of apology laws were endogenous to malpractice trends within a given state, but our results remain consistent to various robustness checks. The analysis finds that the apology laws increase the number of closed claims, particularly among cases involving the most severe medical injuries. One can attribute the increased number of closed claims to several causes. Most notably, the increase can be due to faster resolution time or an overall increase in the number of claims filed. Consistent with our theory, we find evidence that suggests that the overall increase in closed cases is due to faster settlement times, while in the long run, the total number of malpractice claims being brought forth is declining. For example, on the subset of cases involving insignificant injuries—those cases which are most likely to be resolved in the 3 to 5 years of available data—we see a 16–18% reduction in the number of closed claims. When we investigate the impact of apology laws on the duration of cases by the severity of medical outcomes, the hazard analysis results suggest that conditional on the cases resolved before 2009, the cases with most severe injuries settle 19-20% faster in states that have the apology laws relative to states that do not. As for the claim compensation payouts, again conditional on the cases resolved before 2009, apology laws reduce the 5 claim payouts of the most severe cases by $58,000–73,000 per case and the claim payouts of the “somewhat” severe cases by $7,000–14,000 per case. In sum, we find that apology laws induce faster resolution and lower payments for those malpractice cases brought by patients who sustained the most severe medical injuries.2 The remainder of the paper proceeds as follows: Section 1 provides background information on apology laws, in Section 2 we provide a model, section 3 describes out dataset, in Section 4 we discuss our empirical specification and present our findings, and Section 5 concludes. 1 Background of Apology Law As of January 2009, apology laws had been enacted in 36 states, all of which were enacted between 1999 and 2008 (except for Massachusetts, whose law dates back to 1986).3 Table 1 lists all of the 2 We would expect these reductions to lead to similar reductions in malpractice insurance premiums. Data on medical malpractice insurance premiums was obtained from the Medical Liability Monitor Rate Surveys from 1995 to 2005. To make prices comparable, the data cover a typical contract based on coverage for a $1 million per incident and $3 million per year cap. The impact of the apology laws on medical malpractice premiums for three specialties—internal medicine, ob/gyn, and surgery—was also assessed, but these results turned up almost entirely insignificant. This disconnect could be because medical malpractice insurance premiums are highly regulated and thus slow to respond. While total medical malpractice payments were trending downward after 2001, total insurance premiums continued to trend upward. Also, there appeared to be a great deal of inertia in insurance premiums, which often did not change from year to year. Or it could simply be that our data lack the power to identify the differences. 3 The apology laws were initially identified using a search in lexis-nexis of state legal code using the words apology and malpractice. Our list was confirmed from the website of the Sorry Works Coalition, an advocacy group promoting apologies by physicians. 6 state legal codes pertaining to medical apologies.4 Unlike other tort reforms, which have predominantly been a Republican issue (Durrance, 2009), apology laws are not disproportionately supported by any particular political party.5 While there is no hard data to substantiate the following claim, it is anecdotally accepted that apology laws have been passed due to activist pressure rather than systemic changes in the litigation environment, which means that apology laws are unlikely to be correlated with other changes that affect litigation (e.g., tort reform).6 Crucially, since studies such as Born and Viscusi (2005) and Baker, Born and Viscusi (2009) find significant effects of tort reforms—and non-economic damage awards in particular—affect malpractice payments, we find that apology laws are not significantly correlated with other tort reforms, specifically noneconomic caps, punitive caps, laws on full information disclosure, joint and several liabilities, and collateral source rules. [Insert Table 1 About Here] In a survey of hospital risk managers, Lamb et al. (2003) find that while 92% of managers respond to mistakes with an explanation, and 87% would initiate an investigation, only 68% would include an apology, and 33% accept responsibility. State apology laws have been enacted in order to 4 California, Massachusetts, Florida, Tennessee, Texas, and Washington have general apology statutes that apply across all industries while the other 30 States have specific laws that only protect the statements of apology made by health care providers. The states can be first divided into two types depending on the applicability of these laws: general versus health practitioners only. We use the specification in Table 3 but we create two dummies for the general laws versus healthcare-only laws. We then perform an F-test checking whether we can group the general versus healthcare-only laws together, the F-test fails to reject the null hypotheses that these two types of apology laws have the same impact on claim frequencies and claim severity. Therefore, for the remainder of the paper, we are not going to differentiate between general and healthcare-only apology laws. 5 In regressions not reported in the current paper, we find that political composition in the State Senate and State House has no significant explanatory power on the passage of apology laws. 6 See, for example, the efforts of the Sorry Works Coalition. 7 increase these frequencies. State apology laws are very similar to one another as they tend to be copied from similar templates. Connecticut’s apology law is a typical example. The Connecticut law states that: In any civil action brought by an alleged victim of an unanticipated outcome of medical care, or in any arbitration proceeding related to such civil action, any and all statements, affirmations, gestures or conduct expressing apology, fault, sympathy, commiseration, condolence, compassion or a general sense of benevolence that are made by a health care provider or an employee of a health care provider to the alleged victim, a relative of the alleged victim or a representative of the alleged victim and that relate to the discomfort, pain, suffering, injury or death of the alleged victim as a result of the unanticipated outcome of medical care shall be inadmissible as evidence of an admission of liability or as evidence of an admission against interest. (emphasis added) Depending on the state, there is a slight variation in the types of statements that are protected by these statutes. Protected statements typically include a combination of apology, fault, sympathy, commiseration, condolence, compassion, and admissions of mistakes, errors, and liability. In the legal literature, some studies divide apology laws into one of two categories: 1) full apology laws that protect against all types of apologies including those that contain statements of fault, mistakes, errors, and liability versus 2) partial apology laws that only protect against statements of sympathy, commiseration, condolence, and compassion.7 We use the specification in Table 3, but we create two dummies for the full laws and partial apology laws. An F-test fails to reject the null hypotheses that full and partial apology laws have the same impact on frequency of claims and total compensation payout; therefore, in the remaining analysis, we will not differentiate between full and partial apology laws. To date, this paper is the first economic study to investigate the impact of the state-level apology legislation on claim frequency and claim severity.8 Other studies, particularly in medicine and in law, 7 The divisions between full and partial apology laws are arguably poorly defined. A paper by McDonnell and Guenther (2008) reports eight states as having full apology laws, whereas an article by Morse (2009) reports only five states as having full apology laws. 8 Ho and Liu (2011) offers an extension of the research here. Whereas this paper focuses on identifying the effect using difference-in-difference and then using hazard rate analysis to identify speed as the mechanism that drives our results, Ho and Liu (2011) focus on using diff-in-diff-in-diff to decompose the effects of the law by physician and patient characteristics such as age, gender of patient, field of practice, etc. to identify the locus of the effect. 8 have found substantial effects of apologies on litigation. For example, subjects who are given hypothetical situations about health outcomes report that apologies may reduce the subjects’ likelihood to litigate (Mazor et al., 2004; Robbenholt, 2003; Wu, 1999; Wu et al., 2009). Programs in individual hospitals in Pennsylvania (Liebman & Hyman, 2004, 2005) and Tennessee (Kraman & Hamm, 1999) that encourage apologies and disclosure of mistakes dramatically reduced malpractice payments made by the hospitals that adopted these programs. Most notably, Boothman et al. (2009) conducted at the University of Michigan Health Service reported that their per case payments decreased by 47% and the settlement time dropped from 20 months to 6 months since the introduction of their 2001 apology and disclosure program; Kachalia et al. (2010) also study the consequences of the same program and find that the cases are resolved 26% faster at any given time. While the findings associated with hospital-level apology programs are promising, one is uncertain of the generalizability of these results. The reason why the hospitals in these studies decided to pioneer such programs could be endogenous, or there could be other concurrent reforms at the hospital level, as such programs are often implemented by a charismatic administrator. Therefore, the average impact of apologies on medical malpractice litigation would be overstated, leading to concerns about external validity. 2 Theory One reason to formally model the role of apologies in medical malpractice is to clarify the implicit assumptions inherent to the implemented policy. The other is to provide intuition on how the laws will affect the composition of cases that we will observe. We develop the model in two stages. We begin by providing a simple symmetric information model designed to illustrate the intentions of the apology law proponents. In the simplistic model, we assume that the apology laws do not devalue patients’ perception of apologies and level of physicians’ efforts remains constant regardless of the laws’ passage. Under these assumptions, the apology laws increase the likelihood that a doctor apologizes, hasten settlement speeds, reduce total court costs, and unambiguously improve welfare. In the Online Appendix available from the author’s website, we demonstrate that introducing asymmetric information 9 would give us ambiguous prediction of welfare improvement. For example, if physicians’ efforts are unobserved, the laws could lead to increased moral hazard in quality of care. Similarly, the law could devalue apologies and reduce welfare. Let there be two players: a patient/plaintiff (P) and a doctor/defendant (D) who play a game of healthcare provision, apology, and litigation with the following timeline: The doctor chooses effort (e) After observing the patient outcome (h), the doctor decides to apologize or not (a) After observing the psychic costs of litigation (ψl), the patient decides to litigate or not (pl) If the patient litigates, the doctor proposes a settlement offer (S) After observing the psychic costs of going to trial (ψJ), the patient decides to settle quickly or not (pS) The key novelty in this model is that the patient must care about the doctor’s intention in addition to monetary payoffs. If as is standard, plaintiff and defendant care only about monetary payoffs, (Daughety & Reinganum, 2000; Farber & White, 1994; Spier, 1992), doctors would never apologize because patients would not value the apology and would just use the apology as evidence to obtain a higher malpractice payment. We will take the patient’s health outcome, , to be exogenous in order to focus on the ligitation process, and we will assume apologies exogenously shift the patient’s benefits from litigation. In a more complete model health would depend on both the doctor’s effort and the patient’s circumstances. Apologies would serve as a signal of the doctor’s intentions, effectively signaling that the poor health outcome came from unlucky circumstance rather than low effort. We leave the details of such an asymmetric information model to the Online Appendix on the author’s website and Ho (2011) and instead focus on the effect on litigation. After observing the health outcome, the doctor decides whether to apologize ( apologize ( 1) or not 0). The cost of an apology for the doctor is that the apology can be used as evidence against him/her in court. If litigation occurs, since the court cannot observe the doctor’s effort, we 10 assume that the expected judgment, ,9 is lower for better health outcomes and higher if the doctor , apologized since the apology can be used as evidence against the doctor. The benefit of an apology to the doctor is that it increases the psychic cost for the patient to litigate. Numerous case studies suggest that anger is a main motivator for litigation and can overcome the patient’s aversion to litigate (Hickson et al., 1992; Vincent et al., 1994). Studies also find that apologies reduce patient anger, increase communication, and reduce the patient’s motivation to litigate (Liebman & Hyman, 2004, 2005; Ohbuchi, Kameda, & Agarie, 1989; Sloan & Hsieh, 1995). We capture these psychological factors by saying there is a psychic disutility of initiating litigation, disutility for going to trial, , and a psychic . Both disutilities would increase if the doctor apologizes. These psychic costs are modeled as random valued functions of whether a doctor apologizes where 0 for stochastically dominates , . For simplicity, let and be constants for the additional psychic cost of litigating/seeking judgment against a doctor who apologized. disutility follows a uniform distribution such that tendered, Let 0 otherwise for ~ 0,1 1 first order where Assume the psychic 1 if an apology was , .10 be the probability the doctor decides to apologize, which will depend on the health outcome. After the doctor apologizes (or not), the patient observes the realization of his psychic disutility 0 ). It is now the patient’s turn to decide whether to litigate or not. Let be the probability that the patient litigates. If the patient litigates, the doctor will offer a settlement as a of litigating 1 (or take-it or leave-it offer. Let 9 be the probability the patient accepts the settlement offer. If the patient This is the amount that the patient receives and the doctor is required to pay after accounting for the probability that the patient wins. 10 All we really need is that these are random valued functions of whether a doctor apologizes where ψi(1) first order stochastically dominates ψ_i (0) for i {L,J}. However a uniform distribution will make calculations more convenient. 11 does not accept the offer, the patient incurs legal costs and the doctor incurs legal cost , are awarded according to an exogenous function: and damages . To summarize, the patient’s utility depends on the patient’s health plus expected malpractice payments net of litigation and psychic costs, while the doctor’s utility depends on the doctor’s cost of effort minus expected malpractice payments and the economic costs of litigation: , , Π pL , pS (1) , ΠD . Where the malpractice payments net of costs Π and ΠD are given by: ΠP p L , p ΠD h, a , 1– , 1– – (2) . We consider pure strategy sub-game perfect Nash Equilibrium. Solving by backward induction yields the following results: Proposition 1: In the symmetric information pure strategy Nash equilibrium with interior solution, when a doctor apologizes: a. Probability of settlement increases b. Probability of initiating litigation i. may increase or decrease An increase is more likely when the value of apology as evidence is higher , 0 or when the total cost of seeking court judgment ii. ,1 is higher. A decrease is more likely when the psychic cost of litigating a doctor who apologizes is higher. Proof in the Appendix. A rational doctor would only apologize if it maximized her utility. Since we assume the doctor has only pecuniary motives, doctors who apologize have lower expected malpractice payments. Similarly, since apologizing makes it more costly for patients to litigate and to seek judgment, apologies increase the likelihood of settlement and decrease the likelihood that a patient initiates litigation. What may be 12 unexpected is that an apology is more likely to increase the likelihood of litigation when the costs of seeking a court judgment are higher. This result is due to the fact that one cost that deters patients from initiating litigation is the possibility of suffering costly court costs. Since an apology makes settlement outside of court more likely, it also makes it easier for the patient to litigate. Introducing Apology Laws Now suppose that the legislature passes a law excluding apologies as evidence in court. Assume that the law has no effect on how apologies affect psychic costs and that the only effect of an apology is to reduce judgments such that the new expected judgment function, , treats all cases as if no apology was ever tendered: ,1 ,0 , 0 . This section considers the symmetric information case to demonstrate how the law was intended to function, but we will demonstrate how asymmetric information assumptions that endogenize the effects of apologies and the size of judgments change these results in the next section. Proposition 2: In the symmetric information pure strategy Nash Equilibrium with interior solution, after the introduction of an apology law: a. Apologies are more frequent b. The probability of litigation decreases. c. The probability of settlement increases. d. For any particular case, the settlement offers decrease e. For any particular case, net expected malpractice payments f. and decrease Excluding psychic costs, social welfare increases Proof in the Appendix. The law eliminates the primary cost the doctor faces from an apology.11 Thus in a symmetric information world, the law has exactly its intended effects. Doctors apologize more often. Less litigation 11 The only cost that remains is that an apology makes settlements more likely making it easier for patients to initiate litigation since they know court costs are now less likely to be incurred. 13 is pursued. The litigation that does occur is more likely to result in settlements, and when a settlement is reached, payments are smaller. Since the law reduces the number of cases that go to court, the dead weight loss of court costs are reduced, so excluding psychic costs, social welfare increases. A key variable we will consider in the empirical analysis is the time to settlement. We think of as the probability of settling early, rather than prolonging pretrial negotiations and incurring costs , , from a long negotiation process. Thus we will interpret the model as predicting that apology laws should increase the speed of settlement. Propositions 1 and 2 depend on the linearity assumptions and the uniform distributions of psychic costs, however, qualitatively similar results can be obtained using more general functional forms so long as we assume symmetric information. Introducing Private Information While the above results are consistent with the intentions of the law, the analysis presumes that there is no private information between players. Much of the past theoretical literature on malpractice litigation has focused on asymmetric information, and thus, introducing private information is important for increasing the validity of the model. Unfortunately, private information also makes the model’s predictions indeterminate. The impacts of private information are considered in more detail in Ho (2011) as well as in the Online Appendix on the author’s website. Moral hazard and the provision of care: The current model has no moral hazard since doctor effort is symmetrically observed. However, if effort were unobserved, reducing the consequences of poor performance could reduce quality of care (Polinsky & Rubinfeld, 1988). Disclosure of health outcomes: If patients are imperfectly informed about whether a mistake was made, laws that encourage doctors to disclose more information could induce patients to litigate more frequently. Signaling and the value of an apology: The above model assumes an exogenous value of an apology. If instead, an apology provides information about a doctor’s intentions, and the psychic cost of litigating is 14 the probability of suing a good doctor, rather than a negligent one, then apology laws could reduce the signaling value of apologies, and thus increase litigation. Effects on bargaining and settlement: A substantial literature considers the effect of asymmetric information on pre-trial bargaining and the effect of over-optimism (see Spier (2002) for a review). Both literatures predict that increased information disclosure should lead to faster settlements. Unlike the other effects of private information, this literature tends to conform to the law’s intentions. Private information and judicial decision making: The law presumes that less evidence against doctors means more judgments in favor of doctors. However, if a rational risk neutral judge knows that the same level of malpractice is occurring, then the judges should on average award the same level of damages. Summary We are not aware of any comprehensive dataset of malpractice cases that includes information about apologies. Thus we will focus on the theory’s predictions regarding the probability of litigation, the probability of settlement, and the size of claim payout. Ideally, we would like to conduct our analysis on the total number of malpractice claims ever filed (including both open and closed claims) in order to examine the probability of settlement and claim payout. Unfortunately, to our knowledge, comprehensive public data on all open claims that have yet to be resolved also does not exist. This presents some challenges which will be addressed in the next section. 3. Data To assess the impact of these various types of apology laws, we use data drawn from the NPDB Public Use Database (2009). Due to the Federal Health Care Quality Improvement Act (HCQIA), all malpractice payments—either as part of a settlement or as part of a court judgment—made by, or on behalf of, a licensed health care provider must be reported within 30 days. The NPDB contains all malpractice cases with non-zero payments12 and it provides additional information about each claim 12 The NPDB dataset is not free of problems. It has been criticized because of a “corporate shield” loophole, through which malpractice payments made on behalf of a practitioner end up excising the practitioner’s name from the data 15